Skip to main content
Top

2024 | Book

Accelerating Discoveries in Data Science and Artificial Intelligence I

ICDSAI 2023, LIET Vizianagaram, India, April 24–25

insite
SEARCH

About this book

The Volume 1 book on Accelerating Discoveries in Data Science and Artificial Intelligence (Proceedings of ICDSAI 2023), that was held on April 24-25, 2023 by CSUSB USA, the International Association of Academicians (IAASSE), and the Lendi Institute of Engineering and Technology, Vizianagaram, India is intended to be used as a reference book for researchers and practitioners in the disciplines of AI and data science. The book introduces key topics and algorithms and explains how these contribute to healthcare, manufacturing, law, finance, retail, real estate, accounting, digital marketing, and various other fields. The book is primarily meant for academics, researchers, and engineers who want to employ data science techniques and AI applications to address real-world issues. Besides that, businesses and technology creators will also find it appealing to use in industry.

Table of Contents

Frontmatter
Analysis of Fraud Detection Approaches in Online Payment Systems

Detecting fraud is an important topic that is appropriate to many different industries, together with the banking industry, government agencies, insurance companies, law prosecution, and many more. E-commerce platforms already provide a wide variety of different payment methods for use online across the globe. The development of fraud detection algorithms those are both powerful and efficient. When it comes to uncovering fraudulent activity, a wide variety of approaches have been tried and tested. This chapter presents the strategies for detecting fraudulent activity in online systems and analyzes the best-performing methodology in the fraud detection process, as well as discusses the role of unbalanced data handling, such as resampling strategies methods.

Kothapalli Mandakini, Kattula Shyamala
Investigating Context-Aware Sentiment Classification Using Machine Learning Algorithms

In today’s world knowing the opinion of people is a widely used research topic in natural language processing, and context-aware sentiment analysis has emerged as an important research area in recent years. This chapter conducts a survey of context-aware sentiment classification using machine learning algorithms. In this, first, provide an overview of the key concepts and challenges in context-aware sentiment analysis, and then review the existing literature on context-aware sentiments using machine learning algorithms. The chapter classifies the studies depending on the types of machine learning (ML)-based algorithms used, which are supervised, unsupervised, and deep neural network–based learning, and discusses the strengths and weaknesses of each approach. Furthermore, we identify the major challenges in context-aware sentiment analysis using machine learning techniques, such as data sparsity, context feature selection, and model interpretability, and discuss the potential solutions and future directions. We also provide a summary of the applications of context-aware sentiments using ML techniques in different domains, which are marketing, customer service, restaurants, politics, and health. This survey gives a detailed overview by considering present state-of-the-art context-aware sentiment classification using machine learning algorithms and highlights potential opportunities along with challenges in this field.

P. Ashok Kumar, B. Vishnu Vardhan, Pandi Chiranjeevi
Opioid Recommendation to Arthroplasty Patients, Using Pearson Correlation and Shapiro Wilk Test

Opioid abuse, overdose, and drug addiction have become public health issues because of the sharp increase in both prescribed and over-the-counter opioid use. The opioid overprescription or the treatment of both acute and ongoing pain, where a 36–1 size strategy is routinely utilised yet can have significant effects for those who take one unusual dose, is directly related to addiction and overdosing. The pattern for each patient should be taken into account to reduce overprescribing and overdoses. This study provides a model for predicting all patients’ respective levels of opioid intake, by using machine learning in the first 2 weeks after discharge given that patients who undergo total joint replacement surgeries are usually advised to take opioids. Data from consumer surveys, patient prescription histories, and electronic health records are gathered to look at the extent of short-term opioid use following joint substitution procedures. The surveys do, however, contain a sizeable percentage of missing responses, which lowers the quality of the data. A semi-supervised learning approach that uses Bayesian regression to give pseudo-labels is put forth to get around this problem. This algorithm predicts the missing survey responses on the percentage of patients who initially take opioids. Next, to enhance classification performance, false labels are applied to those patients in accordance with the prediction. Numerous tests have shown that the resultant patient categorisation performed better when using a model of semi-supervised learning. We anticipate that by employing such a model, healthcare professionals should be able to modify the dosages of opioids to match individual patients’ actual needs, which will help in controlling pain with prescription opioid management.

R. Menaha, P. Shruthika, A. R. Abdul Ashiq, M. Akshay, V. Santhosh
Sindhi POS Tagger Using LSTM and Pre-Trained Word Embeddings

The chapter shows the development of a part-of-speech (POS) tagger for Sindhi, which is a highly resource-poor language. For our study, we have used Sindhi in the Devanagari script. We have developed a corpus of thirty thousand POS-tagged sentences and developed Glove word embeddings for the monolingual Sindhi corpus of 1 lac sentences. We have used LSTM and Glove word embeddings for developing our POS tagger. The developed system was compared with a Hidden Markov Model (HMM)-based POS tagger developed for Sindhi (Nathani and Joshi, Part of speech tagging for a resource poor language: Sindhi in Devanagari script using HMM and CRF. In: Proceedings of the 18th international conference on natural language processing, 2021). The evaluation shows significant improvement over the previous study. The HMM POS tagger achieved an overall accuracy of 81.85%, while the proposed LSTM tagger achieved an accuracy of 96.15%.

Bharti Nathani, Palak Arora, Nisheeth Joshi, Pragya Katyayan, Shivani Singh Rathore, Chander Prakash Dadlani
An Early-Stage Colorectal Cancer Detection from Colonoscopy Images Using Enhanced Res-UNET

The segmentation of the polyps during colonoscopy is one of the most crucial factors for successful colon cancer diagnosis. Due to the wide variety of sizes and shapes of these polyps, this process can be challenging. The introduction of deep learning models in biomedical image analysis has shown a great impact on disease detection and segmentation. Most of the deep learning models concentrated on larger size polyps only, smaller size polyps may not segmented accurately during the procedure and can cause cancer in 5–7 years. This can have a significant impact on the early-stage detection of colon cancer. This chapter proposes an enhanced Res-UNET-based deep learning–based model that can improve the performance of segmentation on small-size polyps. We evaluated the proposed model on ETIS-LaribPolypDB, CVC-ColonDB, CVC-ClinicDB, Kvasir-SEG colonoscopy image datasets. The results show that the proposed enhanced Res-UNET model achieves the top Dice segmentation accuracy and top Hausdorff distance over the recent state-of-the-art models.

Mohan Mahanty, Bandi Vamsi, Madhavi Dasari, Setti Vidya Sagar Appaji
Prediction of Rice Leaf Diseases at an Early Stage Using Deep Neural Networks

Rice is the major crop in India, and India has been the biggest exporter and the second-largest producer in the entire world, so it is heavily reliant on rice for its economy and food supply. There has been an increase in rice production from 53.6 million tons in fiscal to 120 million tons over the last 40 years. Healthy and proper rice plant growth is required to maintain food production and supply in accordance with people’s needs and demands. The major diseases affecting these rice plants are leaf blast, hispa, and brown spot. The advancement of deep learning paved the way for these diseases to be detected using computer vision. Many researches have been conducted in the detection of rice leaf diseases, where the rice leaves have been deeply affected at a later stage. In this chapter, we propose an effective deep learning model for an early-stage disease detection of rice leaves. The proposed model was developed by an ensemble of the pretrained DenseNet-201 with the Naïve Inception module. The model was trained and tested over a Rice Diseases Image dataset; it attained an overall disease detection accuracy of 87.69% and surpasseed the existing models.

Mohan Mahanty, Bandi Vamsi, Y. Srilatha, Bhanu Prakash Doppala
A Study of the Impact of Implementing a Procedure for Creation of Risk Factor Software

When applied to the worldwide advancement of programming, the ob- stacles of requirements engineering become doable. Something is difficult for a variety of reasons. Chances are you’re one among them because the global im- provement outlook is more receptive to gambling. As a result, it could be one of the primary reasons for taking requirement engineering testing seriously. To begin, it is vital to identify the variables that actually cause these dangers. The factors and threats that these elements may bring about are then separated in this chapter. An orderly writing survey is done in the framework of the worldwide programming improvement perspective for the observable evidence of these variables and the hazards that may emerge during the necessity designing cycle. The list suggests a moderate improvement for aiding exercises in necessity designing in a global programming advancement worldview. This work is very beneficial for those with less experience working in global programming advancement.

Mahanti Sriramulu, Satyabrata Patro, G. Nirmala Joycee, Kondapalli Ashok
Chat Analysis and Spam Detection of WhatsApp Using Machine Learning

In recent years, the WhatsApp app has been among the most effective means of communication. Conversations of various kinds that occur in a group of individuals and are centered on a variety of subjects are included in WhatsApp chats. This web application aims to provide a detailed analysis of chats between users of a particular group and two individuals. Regardless of the topic on which the conversation is based, this web application can also be used to detect spam messages. This web application displays the total number of messages and the graph of the month line, as well as the most frequently used words and the active participants of the conversation. This web application also detects spam messages sent by a person and shows the total number of spam messages in a particular chat. The benefit of this tool is that it has been developed using straightforward Python modules such as Pandas, Streamline, Seaborn, and Word Cloud, as well as machine learning algorithms that are used to create data frames and generate various graphs that are then displayed in the application. These algorithms are effective and resource-light, making them suitable for use with large datasets.

Dornala Teja, Sundige Kiran Kumar, Dulam Mani Chandra, Subhani Shaik, N. Sreevidya
A Study of Hate Speech Detection Using Different Models

This paper focuses on a comparative study of algorithms that can detect social hate speech/mongering in the cyber world. The act of using online platforms to harass or bully others, known as cyberbullying, is a significant problem that has serious consequences for its victims. Hate Speech is one such issue under this category. Hate speech leads to violence, bullying, harassment and disturbances. One of the primary forms is text on social networks, affecting over a billion users. Despite the use of machine learning models, including deep learning, to tackle cyberbullying, the challenge of effectively classifying and monitoring such behavior still persists widely. To tackle this issue, an existing approach involves the use of a novel concept of CNN.The proposed approach is to understand different machine learning models, such as Logistic Regression, Naïve Bayes, Decision Tree, Random Forest and BERT. The use of NLP techniques while training the model can improve its performance in processing and analyzing textual data, the model can better understand the language used in the dataset and accurately classify instances of cyberbullying. The goal is to effectively monitor and prevent cyberbullying conducted in the form of Hate speech text, using machine learning and language processing techniques.

Y. Mihir, T. Srikanth, Marlapalli Krishna, B. S. B. P. Rani
Image Feature Narrator for the Blind

There are millions of visually impaired people in the world. The quality of life of visually impaired people is greatly affected by their inability to interpret visual text and images. For sighted people, the brain can recognize image features, but for blind people, this is not possible. So we create a model to help the computer identify features in the image. This model is a combination of recurrent neural networks (RNN), long short-term memory (LSTM), and convolutional neural networks (CNN). These deep learning algorithms use natural language processing to extract image attributes. Images are recognized using NLP, which uses natural language to describe images. In this chapter, we use CNNs and LSTMs (RNNs) to prepare a feature generator and use the gTTS API to listen to image details (Google Text To Speech). This API allows you to support multiple languages ​​and provide audio at a set rate.

R. Soujanya, Sahithi Amerineni, Jahnavi Chunduri, Sree Akshitha Muktheswaram, Aishwarya Peri
Recognition of Indian Gestural Language Through Neural Networks: Narrative Approach

Gesture-based communication acknowledgment is useful in correspondence between marking individuals and non-marking individuals. Different research projects are underway on various communication via gesture acknowledgment frameworks around the world. The examination is restricted to specific countries as there are country-wide varieties accessible In this paper the hand signals compared to ISL English letters in order are caught through a webcam. In the caught outlines the hand is divided and the neural organizations are utilized to perceive the letters in order. The elements like points made between fingers, the number of fingers that are completely opened, completely shut, or semi-shut, and recognizable proof of each finger contribute to the neural organization. Trial and error were accomplished for single-hand letter sets and the results are summed up. In this paper, we have mentioned a comparison of various techniques for using recognition of gestural language like a neural network, CNN, deep learning, artificial neural network, and autoencoder work done to determine better results.

Ramana Babu Budimure, A. S. Lalitha, Golagani A. V. R. C. Rao, Satyanarayana Mummana, Bosubabu Sambana, Chandra Sekhar Akula
Frequency and Voltage Control of Multi-Area Multisource Power System Using Whale Optimization Algorithm

The stability of terminal voltage and nominal frequency in an interlinked power system (IPS) is a foremost challenge. The active and reactive power requirements are impacted by load fluctuations or other disturbances, which has a negative impact on how well IPS functions normally. In order to keep the terminal voltage and frequency at their recommended levels, Tie-line is used to connect the loops of the automatic voltage regulator (AVR) and load frequency control (LFC). In this work, a control methodology based on a 2DOF-PID and PIDA controller is used in a multi-area IPS in mutual LFC and AVR. The optimal parameters of 2DOF-PID and PIDA controller were obtained using Whale Optimization Algorithm (WOA). The simulation outcomes of WOA-2DOFPID and WOA-PIDA control schemes were matched with recently developed nature-inspired computations. In addition to LPBO-PIPD, AOA-PIPD, and MPSO-PIPD, the suggested control schemes WOA-2DOFPID and WOA-PIDA are clearly more reliable, as evidenced by the comparison of the controllers.

Kumaraswamy Simhadri, B. V. S. Acharyulu, Udiyana Jagadeesh, Karri Dilleshwara Rao, Boddepalli Santosh Kumar, Ahmad Ali
Tumor Prediction Using Microarray Gene Expression Profiles Through SVM and CBFS

The introduction to microarray technology has created a breakthrough in gene profiling which in turn has led to precise prediction of tumors. Microarray gene expression profiles are being used for classification of tumors into cancer-causing malignant and benign types. Furthermore, they are used to identify possible gene markers for each of the types in cases of malignant tumors, thus making a precise cancer diagnosis possible. This study mainly focuses on a supervised learning algorithm—support vector machine (SVM)—the use of various feature selection methods to enhance the precision score A forward greedy search strategy based on consistency and another known as the signal-to-noise ratio metric was used to discover the potential gene markers. According to the experiment findings, the consistency-based feature selection method is very effective for predicting cancer subtypes when combined with ISVMs. This would result in better prediction of cancer types based on the gene profiling data which would serve to be quite useful in the medical diagnosis of cancers.

Ankan Bandyopadhyay, Abhishek Bandyopadhyay, Debasis Chakraborty
Advanced Machine Learning Approaches for Improving Traffic Flow Predictions in Smart Transportation Systems

Traffic flow extrapolation is a vital aspect of intellectual transportation systems, as it facilitates the smooth and efficient management of traffic. However, traditional traffic flow prediction methods have several limitations, including a lack of accuracy and the inability to handle large and complex data. In this study, advanced machine learning techniques are suggested for enhancing the precision and efficiency of traffic flow predictions in intelligent transportation systems. We collect and preprocess a large and diverse dataset of historical traffic flow data, and use techniques such as feature engineering, model selection, model evaluation, model tuning, and model ensemble to develop a robust-based traffic flow prediction system. The system suggested in this research integrates KNN, Random Forest, and ARIMA machine learning algorithms to forecast traffic flow considering various factors including weather, road conditions, and traffic volume. The outcomes show that the projected system significantly enhances the accuracy of traffic flow predictions compared to traditional methods and can handle a wide range of traffic scenarios and conditions. This research demonstrates the potential of advanced machine learning approaches for improving traffic flow predictions in smart transportation systems.

S. Kanakaprabha, G. Ganeshkumar, A. Sureshkumar, T. Udhayakumar, M. Nisha, P. Poornaprakash
Deep Learning–Based Surrounding Descriptor for the Visually Challenged

The proposed work presented in this chapter, “Deep learning–based surroundings descriptor for the visually challenged,” aims to make a system software that helps visually impaired or blind individuals in perceiving and understanding their surroundings. Visually challenged persons are not able to enjoy the surroundings and nature as we normally do, and this project aims at improving their quality of life. The system will capture images to detect environmental features and then provide real-world auditory feedback to the user. The goal of this project is to create a reliable, user-friendly, and assistive technology and an affordable device software that improves the visually challenged person’s life and makes them independent, allowing them to move around more safely and confidently. This project uses several deep learning and NLP technologies such as CNN and LSTM. The image inputs are taken and the features are extracted using the VGG16 model, which is an advanced version of CNN that is great at object identification and localization. In addition, we used LSTM for training the model with the extracted features and the corresponding captions. Finally, when a user gives an image input to the trained model, it predicts the caption, and then it converts the text-to-speech for the user. The Surroundings Descriptor technology aims to solve the difficulties that visually impaired people face when navigating their surroundings and to improve their experience of life. The project includes technology design and development, system testing and evaluation, and improving the model based on user feedback. Finally, the Surroundings Descriptor has the potential to significantly improve visually impaired individuals’ mobility and independence, allowing them to participate better in public life and live more fulfilling lives.

T. Mohana Naga Vamsi, K. Ravi Shankar, G. Karthik, M. Puruvi, B. Chanti
Wind Power Prediction Using Artificial Neural Network Model: A Case Study

Considering the high level of pollution that threatens our earth, energy from the wind represents a major alternative to fossil fuels, thanks to its high potential and large production. In order to benefit from this renewable energy, several methods of predicting the production of wind energy have been created and applied. In this article, a case study of wind energy forecasting for a Spanish wind installation in Sota Vento is realized. The employed methodology uses artificial neural networks based on a feed forward algorithm, a commonly applied method of artificial intelligence. The tool used is the NN-toolbox from MATLAB, and the model chosen is the Nonlinear autoregressive exogenous. After the model simulation in MATLAB, the results show that the determination coefficient R has a value close to 1 and the mean square error tends to 0. This confirms the average prediction of the model. For better performance, it is preferable to input more historical data and to combine ANNs with metaheuristic algorithms.

Doha Bouabdallaoui, Touria Haidi, Faissal Elmariami, Mounir Derri, Ali Tarraq, Meriem Majdoub
Optimizing Wind Farm Design by Incorporating Wind Turbines of Diverse Hub Heights Through PSO

Optimizing the architecture of a wind farm is a realistic approach for increasing its power output. Previous research has suggested that the implementation of turbines featuring varying heights and elevations may enhance the generation of power. However, there have been few studies that use particle swarm optimization (PSO), and the goal was to enhance the park’s configuration, taking into account various hub heights. The authors of this paper investigate the impact of using turbines with diverse hub elevations on the production of a small park. They use PSO to create an objective function based on the Mossetti cost function to reach the turbines’ ideal arrangement while accounting for energy production costs. The findings show that deploying wind turbines with varied hub heights can boost electricity generation.

Mariam El Jaadi, Touria Haidi, Abdelaziz Belfqih, Ali Tarraq, Atar Dialmy, Zineb El Idrissi
Use of Regression Algorithm for Bike Ride Sharing Demand Projection

Bike-sharing systems are becoming an increasingly popular means of transportation in areas of urban populations around the world. Accurate prediction of bike sharing demand is critical for the efficient management of these systems, as it allows for the provision of resources and the optimization of the user experience. Machine learning models have emerged as a powerful tool for bike-sharing demand prediction in recent years. These models can consider a range of factors, including weather conditions, time of day, and the location of the bike-sharing station, in order to generate accurate projections of demand. In this research chapter, with the help of machine learning techniques, the data sources and pre-processing steps through evaluated metrics were used to assess the performance in terms of hyperparameter tuning techniques. Several regression algorithms such as linear, lasso, random forest, gradient boosting, XGBoost, and lightweight regression evaluated performance on the training data and selected the one that performed best. Out of all, XGBoost regression achieved greatly with 88.81% accuracy. Further, it had evaluated that through this experimentation, root means square log error (RMSLE) was a suitable evaluation metric for bike sharing demand prediction since it was well-suited for continuous, positive-valued targets and was sensitive to the relative size of the errors. Thus, this estimation provided a valuable resource for researchers and practitioners interested in the use of machine learning for bike-sharing demand forecasts.

Husain Korasawala, Satyajit Pangaonkar, Reena Gunjan, Prakash Rokade
Quantifiable Procedures for Covid Improvement Expectation Cox Regression Model

Covids are encompassed RNA infections family of Coronaviridae influencing respiratory, gastrointestinal, hepatic, and neurological frameworks. This study classifies deciding methods into two sorts, to be explicit, stochastic hypothesis numerical models and information science/AI strategies. Information gathered from different stages likewise assumes a key part in gauging. In this audit, two characterizations of datasets have been discussed, i.e., large data from the World Wellbeing Association/Public data sets and data from a virtual amusement correspondence. Estimating a pandemic should be conceivable in light of different boundaries, for model, the effect of natural factors, bring forth period, the effect of isolation, age, orientation and some more. These strategies what’s more, limits utilized for determining are broadly focused on in this work. Notwithstanding, gauging methods accompany their own course of action of difficulties (specialized and conventional). This study inspects these hardships and furthermore gives a bunch of ideas to individuals who are as of now fighting the worldwide Coronavirus pandemic.

G. Sathya Priyanka, S. Rita
Bayesian Optimized Random Forest Classifier for Improved Credit Card Fraud Detection: Overcoming Challenges and Limitations

In the financial sector, credit card fraud is a common issue that costs both people and organizations a lot of money. The ability of machine learning algorithms to automatically identify patterns and anomalies from big datasets has made them a common method for fraud detection. In order to increase the detection of credit card fraud, we suggest an improved Bayesian random forest classifier in this article. By using Bayesian optimization to choose the ideal hyperparameters for the model, we handle the difficulties and restrictions of conventional random forest classifiers. Using an openly accessible dataset on credit card fraud, the proposed method is assessed and contrasted with other cutting-edge approaches. By obtaining an accuracy of 99.6% and an area under the curve (AUC) of 0.99, our findings demonstrate that the suggested optimized Bayesian random forest classifier outperforms conventional random forest and other benchmark methods. We also examine the significance of features in fraud identification in order to show the model’s interpretability. Finally, we discuss the suggested strategy’s drawbacks and potential future developments. Our study aids in the creation of reliable and effective fraud detection tools for the finance sector. The suggested approach can be used as a decision support tool for fraud analysts and detectives and can be applied to a variety of fraud detection issues beyond credit card fraud.

P. K. Rajesh, S. Shreyanth, R. Sarveshwaran, V. Nithin Chary
Early Stage Detection of PCOS Using Deep Learning

A medical illness known as polycystic ovary syndrome (PCOS) af fects hormones in women who are fertile. A delayed or nonexistent menstrual cycle is caused by hormonal imbalance. Women who have PCOS typically struggle with significant symptoms, including weight gain, facial hair growth, pimple, hair loss, and irregular periods. In some cases, these symptoms may lead to infertility. PCOS should be diagnosed and treated as early as possible because it frequently coexists with adiposity, hyperglycemia, and hypercho- lesterolemia. Since there are no direct images and it is challenging to extract the information needed from an image in order to detect PCOS, the current ap proaches and treatments are inadequate for early-stage detection and prediction utilizing numerical datasets.This issue is addressed by this chapter, which proposes a method for early diagnosis of PCOS, which makes use of ultrasound images of the ovaries. Inception V3 and Sequential CNN are two deep learning convolution neural network methods that are used. In order to predict PCOS, the best algorithm, Sequential CNN, is used, and these methods are contrasted. In this work, an image classification system is used to automatically identify PCOS from the input ovarian ultrasound picture.

K. V. Raghavender, T. Pranavi, G. Dharani, G. Sathwika, S. Tejaswi
Challenges and Advancement in Federated Recommendation System: A Comprehensive Review

In the age of information overload, recommendation systems have become essential for individuals and society. Deep learning has been successful in developing high-quality recommendation systems, but limited training data can hinder their effectiveness. Collecting more data may compromise user privacy and security. To address this issue and improve recommendation quality, federated machine learning, and recommender systems have been fused into a new research area called federated recommender systems. While researchers have made progress in this area, gaps still exist. This review aims to (1) identify the common challenges faced by federated recommendation systems; (2) explore different types of federated recommendation systems based on architecture and machine learning models used; (3) conduct a comparative analysis of federated recommendation systems; (4) examine the communication efficiency, tools, and frameworks used for federated recommendation systems; and (5) suggest future research directions for practitioners in federated learning.

Manisha S. Otari, B. Suresh Kumar, Mithun B. Patil
Conditional DCGAN for Targeted Generation of MNIST Handwritten Digits

Creating accurate images of handwritten numbers is a challenging task that has plagued the field of computer vision for years. Traditional neural networks, such as artificial neural networks (ANN) and convolutional neural networks (CNN), have been unable to accurately generate images that mimic the samples in the training set. However, the Generative Adversarial Network (GAN) has shown promising results in generating images that resemble the training set samples by randomly selecting from the latent space. Despite the effectiveness of GANs, the output of GANs is unpredictable, and it can be challenging to regulate them. To address this, Conditional GANs (cGAN) have been introduced, where additional information can be provided to guide the image generation process. The cGAN enables us to adjust the output of GANs by incorporating additional information from the labels. In this study, the cDCGAN approach to train the MNIST dataset along with the labels was used; cDCGAN stands for conditional deep convolutional GAN, where deep convolutional layers help in extracting more features, and the features extracted can be restored more prominently compared to other GAN architecture like Fully Connected GAN, WGAN, etc.In this research, the training of the cGAN model is done with “10” labels, i.e., 0–9, and the authors intend to do further research in the generation of small case cursive handwriting where there are “26” labels of English letters for which the authors need new dataset of handwritten letters.

Samuel Vasamsetti, Vaibhav Chemboli, G. S. S. Shreyas, Srikanth Thota
Predicting Cryptocurrency Price Using Multiple Deep Learning Models

The development of financial technology has led to the emergence of a brand-new kind of asset known as cryptocurrency. In general, several cryptocurrencies are present all around the world, but Bitcoin (BTC), Litecoin (LTC), and Ethereum (ETH) are considered as best cryptocurrencies. A recurrent neural network (RNN) is best suited for forecasting the prices of several cryptocurrencies. The models offer reliable forecasts based on the mean absolute percentage error (MAPE). In the proposed work, we want to show how three RNN models—the gated recurrent unit (GRU), long short-term memory (LSTM), and bidirectional LSTM (bi-LSTM) models—perform while also examining the MAPE percentages. According to our experimental data, GRU outperforms the other two cryptocurrencies with a very low error rate. As a result, GRU is the model of choice for making more accurate and timely forecasts about the price of cryptocurrencies.

Poodi Venkata Vijaya Durga, Gudala Anusha
Efficient Object Detection, Segmentation, and Recognition Using YOLO Model

This chapter presents a study on efficient object detection, segmentation, and recognition using the YOLO (You Only Look Once) model. The YOLOv3 algorithm is used for object detection and recognition, while contour segmentation is used for object segmentation. The study includes experiments on various images, and the results show that the proposed approach achieves high accuracy and speed in detecting and recognizing objects. The contour segmentation technique also provides precise segmentation of objects in the images. The study demonstrates the effectiveness of the YOLO model in object detection, segmentation, and recognition, and its potential for real-time applications.

Anshuman Sharma, Siddharth Swarup Rautaray
Music Genre Classification Using XGB Boost

It is a scientifically proven fact that music can heal a person. Today, around the world, many individuals are coping with their health issues using music therapies. Music is a part of our lives. Every occasion is covered by our songs, which are categorized by genre for simple access. Songs appeal to us for a variety of reasons, including their melody, words, pitch, feelings, and more. A genre is attributed to a song based on all these characteristics. So, to recommend music knowing which genre it belongs to is very essential. The authors created a system just right for that. Our system will help in classifying a song into various genres available. This was achieved by us using the XGBoost algorithm. Model construction and data preparation are the two procedures implemented. As a result of using the GTZAN (G. Tzanetakis and P. Cook) dataset, which comprises 1000 audio files in 10 genres, each lasting 30 s, has been implemented in our model. With the help of the LIBROSA package, we were able to extract the 14 features and use them to enhance the data when training our model for greater accuracy. As a result, we were able to reach 81% accuracy. Since it offers low data loss and is ideal for medium-sized data like ours, XGB boost has been chosen.

D. Dakshayani Himabindu, Koona Avaneesh, M. Alekhya Mudiraj, Suram Manikanta Reddy, Manthena Sujith Varma
Semantic Web 3.0 Streaming-Based Music Application

Artists have been facing many challenges in the present music industry. Royalties refer to getting the permission to use music from someone who has the right to control how that music is used. There’s no transparency in how these royalty distributions is calculated. This is a serious problem for the musicians; they are often the last to receive any profits, although they are the first to put in the work. Streaming services like Spotify, YouTube Music, Apple Music are a great source for young artists to experiment and assess their talents. But these platforms fall short in several important ways that are limiting an artist’s potential and the revenue that they can make. The research papers cover crypto currencies, decentralized applications, and smart contracts in first and second generations. In third generation, apps are linked to autonomous societies which increased complicated settings in decentralization, and this even delivered information about the healthcare. In our project, we build a Next.js application with a Solana backend, using Anchor Framework. Users will first be required to log in through their Phantom Wallets. Once the wallet connection is established, they must pay 0.1 sols to be able to access the website. Once the transaction is successfully completed, the users of the application may listen to music, make playlists, and follow artists. An artist can upload their music through the upload music option present on the website. Here they enter the song title and the download URL or the cloud link for that song. Each time an artist uploads their work, a block of the song is created on the Solana block chain. The uploaded songs will be visible to all users who follow the artists.

Alladi Abhinav, Venkata Sree Varunya, N. Umesh Chandra, G. Raja Ramesh, Subhani Shaik
A Novel Approach for Text Classification Using Feature Selection Algorithm and Term Weight Measures

Text classification is a method for determining the class label of an unknown textual document. In text classification, the vector representation of a document plays a crucial role in enhancing the efficiency of classification process. Several approaches of text classification use content-based features like words for document vector representation. Words with high distinguishing capability increase the performance of the text classification. Therefore, recognizing such words from a huge number of words is an essential step in text classification. This problem of high dimensional is solved with the help of feature selection methods. In the literature, several feature selection methods are proposed by the researchers based on the information of term distributions in various classes of dataset. In this chapter, we developed an approach for text classification (TC) by combining feature selection algorithm (FSA) and term weight measures (TWMs), in which a new feature selection method is developed to delete redundant features and for selecting relevant features. The recognized features are utilized for expressing the documents as vectors. The value of term in representation of vector is calculated by using TWM. In the proposed approach, a new Term Weight Measure is developed and compared the performance of proposed TWM with several well-known TWMs. Six different classification algorithms namely support vector machine (SVM), decision tree (DT), Naïve Bayes (NB), k-nearest neighbour (KNN), logistic regression (LR), and random forest (RF) are used for generating the model for classification. The experiment is performed on six benchmark datasets in the field of TC. The results showed that the proposed approach showed best accuracies for TC on six datasets compared with different works in the domain of TC.

Ravi Kumar Palacharla, Valli Kumari Vatsavayi
Content-Based Music Video Recommender System Using Cosine Similarity

A recommender system works on the basis of machine learning. The proposed work builds a music video recommender system, and it consists of two phases. First, this work constructs a network that reveals relationships between song names and artists. Second, by using textual feature extraction, link prediction is carried out by considering the song’s name prediction, which uses cosine similarity and the content of the song’s lyrics. The recommendation results for songs are based on an item matrix that uses a recommendation engine trained on a cosine similarity algorithm. The generation of recommended songs makes links to regular expressions (regex) by using a song’s name, its artist’s name, and YouTube’s application programming interface (API). It also uses web scraping to find out more about the links that are based on that request of regexes through YouTube API, the uniform resource locator (urllib) module, and a web browser module that plays the recommended songs. The effectiveness of the suggested method is assessed by calculating several variables: accuracy, precision, F1 score, and recall. The real-world “songdata” dataset has been used in this experiment to yield the work’s results.

Shaik Musarrath, K. Subramanyam, Karishma Shaik, Peddi Anudeep, D. Deepthi, M. V. P. Chandra Sekhara Rao
A Comparative Study on Anomaly-Based Network Intrusion Detection System

We are living in an information era where it is not possible to complete a day without information systems. Heterogeneous systems and devices become parts of information processing, transformation, and transmission. Parallelly hackers acquire potential means to intrude into information systems. The war between makers and hackers for information is never-ending. Evolution of Intrusion Detection Systems (IDS) is gaining pace with the help of machine learning and deep learning techniques. In the recent past, numerous attempts have been made to diminish hacker attacks, and the results are published in the literature. In this paper, we tried to offer some knowledge to the research community in the form of a study on anomaly-based network intrusion detection systems using machine learning and related techniques. Previous papers are filtered for a common dataset named CSE-CIC-IDS2018. The dataset is a collection of logs at the University of New Brunswick’s servers, about various DoS attacks. All the selected papers are compared with respect to the performance of the applied algorithms, and the best algorithms for intrusion detection systems are derived. The study reveals factors that affect the performance of algorithms, where preprocessing is a dominant one along with class imbalance handling. This study certainly provides some knowledge on designing better IDS alternatives.

Potti Harshitha, Popuri Sowmya, Madhu Preethi Akula, Lavanya Samineni, T. Anuradha, Anudeep Peddi
A Spatiotemporal Comprehensive Graph-Based Learning for GIF Sentiment Analysis

As social media has grown, more and more people are using short videos in social media applications to share their thoughts and feelings. On the other hand, the semantic gap problem and the sequence-based sentiment understanding problem make sentiment detection in short videos extremely difficult. From the existing work it has been observed that the temporal feature of a GIF has not been considered with due importance. We propose a spatiotemporal comprehensive graph-based strategy to bridge the gap between spatial and temporal features of a GIF. In our proposed method the region in the temporal domain is considered with the region of spatial domain. Using intra-frame and inter-frame feature map, the weight matrix is learned without prior knowledge, with the help of the combination of the image frames and the feature extractor. In order to demonstrate the effectiveness of the suggested spatiotemporal comprehensive graph-based strategy, we conducted extensive experiments with a variety of GIF.

Mousumi Bhattacharyya, Asmita Roy, Sadip Midya, Anupam Ghosh
A Novel Framework for Detection of Objects from Video Using Deep Learning

Detection of objects in the videos is a most challenging task in surveillance, autonomous driving, robotics, etc. Traditional approaches for detection of objects in videos rely on extraction of features and image processing, which requires a significant amount of manual intervention and having complex and cluttered environments. These algorithms in surveillance systems lack accuracy and most work with labeled samples. The algorithms are time consuming and take significant time for object detection. Nowadays, machine learning is being used mostly for detection of moving objects in videos. Objects detection techniques based on deep learning are promising in the surveillance systems, but they are facing some limitations and drawbacks that attract the attention of the researchers. These deep learning-based algorithms require large amounts of computation, delay detection and tracking, and may not work in novel environments or object categories where training data is not available. This paper demonstrates a novel deep learning method for detection of various objects in videos, for e.g., CCTV surveillance systems. The proposed framework is highly effective in surveillance systems because it can detect and track objects accurately and quickly in real time. The recognition of the moving object has been detected using LBPH (Local Binary Pattern Histogram), eigenface, and Fisherface algorithms. All these algorithms are implemented and compared using different parameters such as accuracy, precision, recall, and F1 score. The results show that the LBPH algorithm enhances accuracy than the other two algorithms. The proposed system achieves up to 89% of accuracy for detection of objects using LBPH, which is 8–12% more than the eigenface and Fisherface algorithms.

Yasmeen Kasam Mulla, Kishor Mane
Recognition of Emotion Behind Speech Using Deep Learning RESNET Algorithm

This work aims to develop a reliable and efficient model that utilizes machine learning techniques to classify speech with high accuracy. The model is built using the Python programming framework Librosa, which extracts tonal information from audio and modifies it using the Fourier transform to plot patterns onto graphs. The system uses deep learning algorithms from ResNet to classify speech, achieving an accuracy range of 85–90%. The model is tested with random inputs to evaluate its functionality, response generation, and potential crashes. The proposed system is based on patterns generated by analyzing collected samples and is designed to produce accurate and error-free results. The system combines the starting model with seed neural network layers and dropout layers to improve performance and uses RELU activation in the hidden layer. The method has a far higher accuracy than previous models (between 60% and 75%).

Jagannadha Varma Pinnamaraju, A. V. D. N. Murthy, G. A. V. Rama Chandra Rao, B. Pradeep, B. Niharika
Exploring Sign Language Recognition Methods: An Effective Kernel Approach

Universally, sign language is the widely used mode of communication for hearing-impaired people. Several conflicting investigations on recognition and classification of sign language have been published in scholarly literature. Eminent studies focused on colored-based hands, sensors and Kinect-based techniques, while few others in the recent times demonstrate the use of machine learning and deep learning methods like ANN, CNN, and SVM. Few methods are expensive and resist user friendliness, whereas others are least effective. This research proposes a kernel learning-based approach for recognizing static alphabets solely with the hands. Four vision-based features including local binary patterns (LBP), a histogram of oriented gradients (HOG), edge-oriented histogram (EOH), and speeded up robust features (SURF) are acquired. The features screwed are separately classified using multiple kernel learning (MKL) with the help of support vector machines (SVM). Using a one-to-all implementation strategy, final recognition is decided upon by a vote process. When compared to current approaches, the simulated results are encouraging.

Josyula Sai Manogna, Vaddula Nandini, B. Rama Krishna, Anudeep Peddi, J. V. Satyanarayana, M. V. P. Chandra Sekhara Rao
Identification of Gradient-Based Attacks on Autonomous Vehicle Traffic Recognition System Using Statistical Method

Machine learning (ML) models like CNNs are employed in many computer vision tasks. One such application is traffic signal recognition systems by autonomous vehicles. ML models are widely adopted for such tasks. The result of any wrong decision taken by the adopted ML model during autodrive mode of the autonomous vehicles is disastrous. Adversarial machine learning, a recent topic of study that has gained attention recently, asserts that ML models are vulnerable to attacks. Adversarial attacks are always coming up with new ways to circumvent detection and compromise security. A plethora of methods are involved in detecting adversarial attacks. Statistical methods are powerful and adaptable to unknown attacks. This chapter focuses on using statistical detection methods for identifying the adversarial attacks on autonomous vehicles data set. Statistical methods were used to successfully identify gradient-based attacks FGSM, DeepFool, JSMA, PGD and Carlini & Wagner attacks with minimum distance measure.

Lavanya Sanapala, Lakshmeeswari Gondi
A Sentiment Analysis of Tweets by Using TF-IDF Vectorizer and Lemmatization with POS Tagging

The rapid development of the Internet, social media, and data preprocessing approaches has led to major demands for new technologies, including data analysis. The primary goal of sentiment analysis is to find the sentiment behind the user reviews being given across various platforms. In this regard, we start the discussion on sentiment analysis research and then describe some of the recent works within this domain. This chapter is presented in two stages. For the first stage, we have considered two datasets, consisting of 16 lakh tweets and nearly 6 lakh Amazon reviews (data entities), and they contain six columns, each of which describes the features of our data entities. Because the data were originally unstructured, we subject them to data processing and applied classification models. The second stage of our project includes arrangement, development, and analysis. To normalize the data by subjecting them to parts of speech (POS) tagging and lemmatization; learning algorithms, like linear support vector machine (SVM) classification, Bernoulli Naïve Bayes, and the logistic regression model, are used. Accuracy is measured to assess the performance of the system, via POS tagging+ lemmatization + TF-IDF vectorizer. Logistic regression had the highest accuracy, at 81%, for Twitter, and linear SVM had an accuracy value of 91%, according to Amazon’s data.

Prashanth Saripilli, G. L. Aruna Kumari, Chandra Sekhar Vannemreddy, Kaleelulla Shaik, Saiteja Madishetty
An Overview of Interval-Valued Intuitionistic Trapezoidal Fuzzy Number (IVITFN) and Its Applications Using Fuzzy Logic Techniques

In this chapter, we introduce the intuitionistic trapezoidal fuzzy number (ITFN) and its interval-valued analog (IVITFN), and we examine their characteristics and survey the algebraic operations that may be carried out on pairs of IVITFNs. We also propose a taxonomy for IVITFN. Finally, we evaluate the manufacturing line by using our suggested IVITFN and algebraic operation, and the experimental findings verify the success and viability of our technique.

S. N. Murty Kodukulla, V. Sireesha
A Machine Learning Approach for DDoS Attack Detection in CIC-DDoS2019 Dataset Using Multiple Linear Regression Algorithm

In the context of cloud computing, distributed denial-of-service (DDoS) attacks are malicious attempts to overwhelm a cloud service, network, or an application with a huge flow of traffic from several unidentified sources, rendering it inaccessible to consumers. In cloud computing systems, where many users share the common underlying infrastructure, DDoS attacks are the most common and destructive types of cyber threats. The attackers’ motive is to hinder and halt the services of applications by posing a stream of illegitimate traffic which is achieved with the help of botnets also known as compromised computers that are infected with certain types of malwares such as viruses or trojans. These botnets can be controlled by the attackers, and the malign traffic is generated by various techniques like IP spoofing to make it more challenging to identify the source of the attack. DDoS attacks can be carried out for a variety of reasons, including monetary gain, social or political action, personal grudges, or just the desire to wreak havoc and disorder.The main objective of this chapter is to propose a model capable of anticipating and detecting the percentage of DDoS attacks using multiple linear regression, a dimensionality reduction approach using feature selection by taking the CIC-DDoS 2019 dataset, and compare the classifier result before and after dimensionality reduction using feature selection with multiple linear regression.

Lakshmeeswari Gondi, Swathi Sambangi, P. Kundana Priya, S. Sharika Anjum
Image Fusion with CT and MRI Images for Improving the Quality of Diagnostics

To create a single image with the most information possible, two photographs of same model are combined through the process of image fusion. Many image-processing applications, including satellite imaging, remote sensing, and picture fusion, are crucial. The study of fusing images has developed to support the development of the above applications, and it has since expanded to include research on medical imaging. The merger of magnetic resonance (MR) and computerized tomography (CT) images is accomplished by using a curvelet-based method, which is presented in this chapter. The goal of merging an MR image and a CT image of the same organ is to create a unique image that contains the most diagnostically useful information on that organ. Wavelet transform has been suggested in some attempts to fuse MR and CT images. Given the number of items and curved forms in medical photos and after fusion, the curvelet transform is expected to perform better than the wavelet transform. According to the simulation results, the curvelet transform outperforms the wavelet transform when merging MR and CT images, in terms of visual quality and peak signal-to-noise ratio (PSNR).

Bopparaju Supriya, Jajula Om Sri Keerthi, Peeka Ushasvi Rachel, Usha Manjari, Subhani Shaik
Comparative Analysis of Lung Sac Inflation

Inflated lung sacs are a serious medical condition that may be fatal to people. An infectious agent, most likely a virus or bacterium, is responsible for this illness. According to the WHO (World Health Organization), lung sac inflation is the third leading cause of mortality in India; a virus, or bacterium, is responsible for this illness. Expert radiotherapists are required to read chest X-rays for pneumonia diagnosis. It’s a painful and tough process to breathe due to an illness. As lung sac inflation is a potentially life-threatening respiratory condition, its early detection is of paramount importance. We provide a methodical approach for pneumonia detection that learns from digital chest X-ray pictures to reliably identify pneumonic lungs. The medical community will benefit greatly from this. We compared the accuracy using five different machine learning models namely random forest, KNN (K-nearest neighbors), CNN (convolutional neural networks), and decision tree of machine learning methods. The CNN model achieved an accuracy of 91.8% in general. The purpose of this project is to predict pneumonia and try to improve the accuracy.

M. Harivirat, D. Manisha, N. Shesha Sarathi, V. Kakulapati, Shaik Subhani
Voice Assistant for Driver Drowsiness Detection

One of the main reasons drivers cause traffic accidents is stress and sleepiness. By requiring that workers get adequate sleep preceding operating, drink tea or another caffeinated beverage, maybe take a break when the symptoms of tiredness arise, they can be prevented. Electroencephalograms (EEGs) and electrocardiograms (ECGs) are two of the sophisticated methods used to identify sluggishness. Although this method’s estimation has a high degree of accuracy, it relies on contact estimation and is subject to a number of limitations in driver fatigue monitoring and sleepiness monitoring. The objective is to ensure that driving remains comfortable throughout trips. By estimating the rate of eye closure, this chapter suggests a technique for detecting indicators of tiredness in drivers.This exercise demonstrates the best way to monitor the lips and eyelashes shown in a video that has been stored on a computer. A webcam is mounted in front of a car, and a participant controls the driving simulation system. This allows for viewing how their sleepiness state shifts from alert to weak and to lethargic by watching the footage that was recorded with the use of a camera. The envisioned framework locates the facial portion of the picture recorded from the clip. The purpose of using the face area is to distinguish the mouth and the eyes from within the face area. When a face is found, the left and right eyeballs, followed by the lips, are used to identify eyes and mouths. This system can operate within the boundaries of the mouth and eyes. Whenever the eyelashes are found, estimating power variations in the eyelid determines whether the pupils are unlocked or shut. If the eyeballs are viewed as shut at four continuous edges, the operator is sleepy.

Yashoda Devi Gondi, Hemanth Modani, Rahul Kinthali, Tarun Kumar Varanasi, Sai Kishore Peddisetti, Divya Teja Mirtipati
Prediction Using Sentiment Analysis on Multi-domain Category: Stocks and Politics Using Twitter Repository

Twitter is a microblogging website and a large public database of various people, products, services, and businesses. The process of analyzing one’s reputation in the public eye is called sentiment analysis. The system should be able to analyze the text, including any special characters like emojis, abbreviations, etc., and classify the user’s opinion as negative, positive, or neutral. It should also present visual statistical analysis (graph) for predicting the likely outcome of the event, given the various opinions of individual users on a particular topic (political affairs, stock prices, etc.). This page focuses on various methods used to analyze product reviews, political pundits, and the general public (who can use the tweet format) to determine whether there are many positive, negative, and neutral opinions that help in product, economy, political results, current events, etc.

G. Monika Sanjana, I. L. V. Sai Kumar, S. Nymisha, R. Sai Gowtham, Preeti Nutipalli
Hyperparameter Optimization for Gradient-Boosted Tree-Based Machine Learning Models and Their Effect on Model Performance

Gradient-boosted tree-based machine learning models have several parameters called hyperparameters that control their fit and performance. Several methods exist to optimize hyperparameters for a given regression or classification problem. However, how and to what extent the tuning of hyperparameters can affect model performance is not well understood. Therefore, we investigated the effect of optimizing the hyperparameters for three commonly used tree-based machine learning models on their performance compared to the default. We evaluated extreme-gradient boosting (XGBoost), light-gradient boosting machine (Lightgbm) and Catboost models for predicting the price of food items listed in an online food delivery application. Mean absolute error, root mean squared error and R-squared served as the evaluation metrics. We found that optimizing hyperparameters consistently improved the performance of the models, regardless of the model type. Also, the model rank depended on whether hyperparameters were optimized or not. These findings suggest that the default settings in common tree-based models should not be relied on for real-world applications. An attempt should be made to optimize hyperparameters for a given model to improve model accuracy. In this chapter, we have evaluated the role of hyperparameter optimization on some of the latest and most advanced tree-based machine learning models. The novelty of our analysis lies in the fact that we demonstrate the role of hyperparameter optimization for advanced use cases such as those involving Catboost-like algorithms.

S. A. Rizwan, V. Deneshkumar, K. Senthamarai Kannan
Medical Data Security with Blockchain and Artificial Intelligence Using SecNet

In today’s IT systems, data is extremely important. Intellectual property, sensitive consumer information, or corporate strategies could all be at stake. We must safeguard the data against hackers. Governance, discovery, protection, compliance, detection, and reaction are the six areas to consider. We need policy, classification, catalogue, and resilience to regulate data security. We must determine where the data is coming from, such as databases, files, and network security. We must safeguard the data using encryption, key management, access control, and backup. We must follow compliance-related reports and keep records and have a capability to detect threats through monitoring, analytics, and alarms. Finally, respond to data detection using cases, automated, dynamic playbooks. Records in cyberspace are distributed ubiquitously and organized by various participants who are unable to trust one another, and the utility of the records in composite Internet is perplexing to allow or authenticate, making it very difficult to permit data sharing on the Internet for mass data and artificial intelligence (AI). With big data handling capabilities enabled by extended AI technologies, more data sources can be secured and made private by incorporating two important components: (i) creating big data with the reliable sharing of data on a wide scale with the assurance of ownership using blockchain technology for reliable data distribution and (ii) forming intelligent security rules using AI technology for secure and reliable data distribution.

P. Laxmi Kanth, O. Sri Nagesh, V. S. S. P. L. N. Balaji Lanka, P. Ramamohan Rao
A Regression Analysis of iPhone 11 Pricing Factors

In today’s competitive market, consumers are always looking for the best deals and discounts before purchasing expensive products such as the iPhone 11 on Amazon. Predicting the price data of such products can help consumers make better decisions about when to buy and how much to save. In this chapter, we aim to predict the price data of the iPhone 11 on Amazon using machine learning algorithms. We collected historical price data of the iPhone 11 on Amazon and used it to train our model. We tested our model on the latest price data and achieved good accuracy in predicting the prices. Finally, we propose a tool for consumers that uses the predicted prices to help them make informed decisions about saving and purchasing the iPhone 11.

Karthikeyan Umesh, G. Nikhileswara Reddy, K. V. Charan Reddy, S. V. Bhaskar, K. R. Tejas
Classification of Alzheimer’s Disease Using Deep Learning Methodologies on MR Images

Alzheimer’s disease (AD) is a progressive neurological disorder. In India, there are over four million dementia patients. Given that dementia affects at least 43 million individuals worldwide, the condition is a global health emergency that requires attention. The elderly are frequently affected by this illness. It influences cognitive cells, which leads to serious health concerns that are related to memory. As a result, for prompt treatment and improved patient outcomes, early and correct AD diagnosis is essential. Recently, deep learning methodologies have brought about positive developments to diagnose AD in the light of the quick advancements in neuroimaging techniques. The goal is to use convolutional neural network (CNN) architectures to create models (custom-built CNN Model, MobileNetV2, and DenseNet169) that aid in diagnosing the illness. A data augmentation approach is used to address the issue regarding the limited size of the Kaggle dataset. The stage of AD can be categorized based on the output of the CNN models. After analyzing the proposed models using evaluation metrics such as accuracy and precision, the accuracies of the custom-built CNN model, MobileNetV2, and DenseNet169 are obtained as 94.35%, 80.49%, and 73.07%, respectively.

N. Rajasekhar, S. Shoban Kumar, Samudrala Karthik, Dundi Rajesh, Rithik Barsal
Taxonomy of Intrusion Detection and Its Effectiveness in Internet of Things

Intrusion detection systems (IDSs) provide defense against cyberattacks in distributed systems. In this chapter, the research studies related to intruder detection technique in IoT are analyzed for privacy and efficiency. Intrusion detections are classified into two categories: (1) Anomaly detection and (2) signature detection. The specification detection method belongs to anomaly detection based on a rule-generating technique. This study shows that anomaly detection technique has notable privacy features compared to the signature technique. The random forest rule generation method offers high privacy and is more efficient in IDS. IoT devices are usually resource-constrained devices and thus require lightweight methods for privacy protection. The feature selection method such as the weighted cuckoo search algorithm has a higher detection value in the IDS.

C. Ramakrishna, Sunke Srinivas, Kumbala Pradeep Reddy, Ranjith Kumar Rupani
Design and Evaluation of Plant Leaf Disease Detection Based on the CNN Classification System

To increase agricultural growth, there is a requirement for the detection of plant leaves in the starting stage. In agriculture, disease detection in plants is an important task and having diseases is quite a common thing. To identify the diseases in plants, a continuous observation is needed, which consumes time and necessitates significant human efforts. Programming technique is used to solve this problem and make it as simple as possible. This chapter proposes the design and evaluation of plant leaf disease detection based on the convolution neural network (CNN) classification system, which contributes to a safe, accurate, and reliable system for disease detection in leaves. Here, k-means clustering algorithm and Gray Level Co-occurrence Matrix (GLCM) are used for feature extraction. The CNN Classification technique is used for leaf disease detection, and the accuracy rate is calculated. The leaf disease detection based on CNN presented in this chapter is compared with existing methods and parameters in the result analysis. This system differentiates diseases in plants and provides accurate classification and compares it with existing methods.

C. Ramakrishna, S. Joy Kumar, N. Venkatesh, Kumbala Pradeep Reddy
Air Pollution Forecasting Using Deep Learning Algorithms: A Review

The emissions from vehicles and factory exhaust and other forms of pollution have caused severe damage to the environment and to people’s health as the rate of urbanization and industrialization has increased at a rapid pace. To safeguard ecosystems and human health, air pollution must be mitigated and prevented. Predicting the future concentration of contaminants in the air can provide more trustworthy data on air pollution. Focusing on fundamental and difficult tasks like preventing and reducing air pollution and addressing extraordinary ecological challenges is essential and difficult now. To effectively predict future air pollution levels, machine learning is a promising method in the field of environmental modelling. In this context, we investigate the state of the art in air pollution forecasting, analyse our findings, provide a critical evaluation of the literature and offer some recommendations for the way forward.

Ravva Ravi, Nalam Sowjanya Kumari, P. S. S. Geethika, Koduganti Venkata Rao, Marada Srinivasa Rao
Health-care Monitoring System Using Artificial Intelligence for Diabetic Skin Diseases

It is projected that computer-aided diagnoses using artificial intelligence (AI) with deep learning (DP) technology will assist in early disease detection to decrease the appearance of diabetic skin diseases. This research proposes a health-care monitoring system using AI to find the hidden problems in diabetic skin diseases that are not similar to other skin diseases. AI-based methods are introduced for detailed research, which determines the outcomes using advanced screening systems on dermatology images of humans. Researchers can also evaluate the current state of this field’s development and identify better future directions that might be studied. Our present research study demonstrates the workflow discussion within the three stages: diabetic skin diseases analysis, dermoscopic image analysis using preprocessing methods, and enhancement of screening system approaches using multitask learning (MTL) technique in AI with deep convolutional neural network. The proposed model has achieved the best and highest results with accuracy of 96.67%, specificity of 96.30%, and sensitivity of 92.31% in the proposed dermoscopic diagnosis systems. In the future, more advanced techniques are needed to understand and identify all types of hidden diabetic skin diseases.

Brahmaji Godi, B. Krishna, B. J. M. Ravi Kumar, Appala Srinuvasu Muttipati, P. V. S. N. Murthy, P. Venakta Uma Krishna Bharadwaj
Proposed Framework for a Doctor’s Appointment Using Chatbot in Tanzania

More than half of the people living in Ukonga, Tanzania, have a hard time getting to modern healthcare facilities, especially in urban areas. The high cost of healthcare and a lack of qualified medical professionals are primarily to blame for the alarmingly high maternal and child mortality rates. Additionally, getting a health-related consultation can be extremely challenging in some remote areas. An innovative healthcare chatbot utilizing natural language processing (NLP) and artificial intelligence is proposed to address these issues. This chatbot can communicate with users via text, allowing them to get accurate and reliable health information, understand the signs and symptoms of various diseases, and even schedule an appointment with a specialist or doctor. The healthcare chatbot has the potential to improve health outcomes and decrease maternal and child mortality rates in Ukonga, Tanzania, by providing access to healthcare information and services in this manner.

Michelle D. Chatumba, Raju Yadav Kumar, Gaurav Jain, Samuel Getachew, Yashpal Singh
Blockchain-Based Ether Transmission That Gives Ledgers Transparency and Security

The absence of security, supervision, and transparency in the Indian government and corporate organisations has a significant negative impact on the willingness of sponsors and organisations. Misusing the funds and presenting false documents to higher authorities was a straightforward process. Blockchain technology can be suggested as the ideal approach to prevent these problems. The shortcomings can be overcome by the aforementioned technologies by making the data decentralised, transparent, and secure. A web application that validates the transaction will be included in this project, and only when the validation is successful on both ends will the data be transmitted into the chain and the individual added to the chain. In this scenario, unauthorised transactions or fraudulent records would not be permitted to enter the chain. As a result, this initiative can guarantee the transactions’ transparency and trustworthiness.

T. Chalapathi Rao, Krishmani Nayak, P. Anjaneyulu, T. Madhuri, S. Rajesh, S. P. V. Dedipya, P. Sai Sharma, P. Aditya Reddy
Fruit Identification and Classification Using Machine Learning

The most important factor in selecting fresh fruits is their identification and quality indication. We cannot inspect every fruit since it would take too much time and effort, and we always want to buy the freshest fruits when we go shopping. Fruits can get harmed, rotten, and impacted by their environment. With the aid of image processing and machine learning, we can recognize fruits and classify them into different classes, making it simple for anyone to choose the fresh fruit available. In this chapter, we offer a useful technique for classifying and identifying fruits. We used supervised learning to train the model for classification. With the help of Keras sequential model for multi-class classification, we implemented the CNN. Nonetheless, because of the similarities in colour, shape, and size of fruits, researchers continue to have difficulty in classifying them. By creating a good model for the identification and classification of fruits, this effort aims to address some of the difficulties encountered by the earlier researchers.

Vikas Cherala, Kiran Nakka, Gayathri Enugula, K. Vigneswara Reddy, Sunil Bhutada
Impact of COVID-19 on Society and Review of Machine Learning Algorithms in Diagnosis

The goal of this work is to examine the ways in which machine learning techniques and applications are used for COVID-19 research and other endeavors. More attention has been paid by authorities and researchers to fundamental statistics and epidemiological methods than traditional methods to predict the international COVID-19 epidemic. One of the biggest obstacles to stopping the development of COVID-19 is insufficient and incomplete medical tests to detect and find treatment. To solve this problem, some statistical-based improvements are deployed leading to a partial solution up to a certain level. Machine learning has employed a variety of tactics based on intelligence, methods, and tools to solve problems in the pharmaceutical industry. This work explores how innovative frameworks such as machine learning respond to the difficulties brought on by the COVID-19 epidemic.

S. Sivaramakrishnan, Kiran Kumar Bonthu, G. Hariharan, J. B. Amarjith, J. Poorvi, Adik Thomas
A Review of Mango Fruit Disease Detection and Pesticide Suggestions

The mango tree is susceptible to a number of illnesses during its whole life cycle. Disorders in mango fruit are a significant source of economic losses and loss of production in the agricultural business worldwide, particularly in developing countries. This chapter presents a method for detecting and classifying mango fruit diseases that have been experimentally validated and demonstrated to be effective. The following actions make up the image processing–based technique that has been suggested as a solution: The first step in the process involves using the K-means clustering technique to segment an image; the second step involves extracting some features from the segmented image; and the final step involves using a support vector machine and a k-nearest neighbor, respectively, to classify the images into one of the classes. Our experimental results demonstrate that the proposed system can significantly improve the accuracy with which mango fruit diseases are detected and classified and their classification automatically.

Vigneswara Reddy K., A. Suhasini, V. V. S. S. S. Balaram
Deep Learning Approach for Expression-Based Songs Recommendation System

Both music fans and consumers may appreciate music, which is a great method for individuals to express themselves. With the development of technology, there are more musicians and more people who enjoy music, and this raises the challenge of manually selecting music. Song suggestions have been around for a while, but in most cases, they are created after learning about the user’s tastes over time. For example, they may take into account the user’s prior song selections, how often he listens to music, and other factors. In this research, we suggest a novel method of song suggestion in which a person’s mood is predicted from his or her photo, and songs are then suggested that most closely match the mood indicated. FER 2013 was used to train this model. Images of faces in greyscale measuring 48 × 48 pixels make up the data. As a result of the faces being automatically registered, each picture has a face that is about in the centre and takes up nearly the same amount of area. Each face must be assigned to one of seven categories, with zero denoting anger, one disgust, two fear, three happiness, four sadness, five surprise, and six neutral. A total of 3589 examples make up the public test set, whereas 28,709 instances make up the training set. We are employing Resnet50 in place of VGG16 and normal sequential CNN. After building the model, we are fine-tuning it again to get more accurate results.

Pasumarthi Koushik, Patha Ruchitha, Devarapally Adithya, Subhani Shaik, Sunil Bhutada
System Sound Control Using Gesticulations

Artificial intelligence is a field of computer science dedicated to solving reasonable problems mostly associated with human intelligence such as pattern recognition and problem solving. This chapter proposes a project controlling the sound (increasing or decreasing the volume) of a computer or a laptop using real-time hand gestures as input through a web camera, which is not necessarily a high-resolution camera, which makes it a cost-efficient project. Making use of gestures is one of the leading aspects contributing toward artificial intelligence and development in the field of technology. This project is implemented using the Python programming language. Here, we make use of the OpenCV library of Python to read the gestures from the user through the web camera. OpenCV is a widely used tool for processing images. To perform hand recognition and hand processing, we make use of the Hand Landmark Model of the MediaPipe Hands library of Python. To control the audio or volume of the system, we make use of the PyCaw Library of Python. Using these Python libraries makes it easy to implement complex projects as most of the required operations or functionalities can be done using a pre-defined function in the particular library.

Mohammed Uzair
Modified Holographic Ricci Dark Energy Cosmological Model in Modified Theory of Gravity

This work analyses a collection of solutions representing the modified holographic Ricci dark energy (MHRDE) cosmological model in modified f(R, T) gravity within the framework of the universe’s late-time accelerating expansion in an anisotropic axially symmetric cosmological model with pressureless matter substance. To solve the field equations, Chen and Jing’s modified holographic Ricci dark energy, the hybrid expansion law, and a relationship between metric potentials are all used. The EoS, matter-energy density, anisotropic dark energy density, and deceleration cosmological parameters have been calculated. We looked at a few important features of the derived model to see if they were consistent with the recent findings.

P. E. Satyanarayana, K. V. S. Sireesha, N. Sandhya Rani
Web Application for Banking Churn Prediction Using ANN

Customer in action or disengagement over a period of time is referred to as churn. In this project, we are developing an application that will automatically deliver offers to customers who are about to depart when it has been determined that they are leaving. With this, we have finished classifying various algorithms and identifying algorithm churn. To create the web application, we used 50,000 customers’ data and 29 distinct parameters. This project’s primary goal is to identify high-risk departing clients. We use an artificial neural network to solve the problem. We have discovered 98% accuracy and 0.1 loss as a consequence. The maximum accuracy is therefore discovered, and neither an overfitting nor an underfitting model is discovered. Additionally, an accurate forecast is discovered, allowing the bank to deliver offers to the consumer and lowering the churn rate. As a result, bank profit is increased. Python is being used to implement this project. Notification is sent by a telegram.

Sandhya S. Kharat, Charushila V. Rane
Providing Security Properties of Cloud Service by Using REST APIs

REST APIs can be used to programmatically access the most recent cloud and Internet services. As demonstrated in this paper, an inducer should be able to compromise a service by taking advantage of vulnerabilities in its REST API. Four security principles are discussed here. Desirable features should be able to be captured through REST APIs and services. The four security principles are as follows: X, X, X, and X. Examples of HTTP requests include GET, PUT, POST, and DELETE. Roy Fielding developed the idea of representational state transfer, and his work on representational state transfer is known by the abbreviation REST. A nice illustration is the usage of a POST request to create a record and a GET request to retrieve it. The first is an update request (PUT), and the second is a deletion request (DELETE). A web architectural design called REST controls how both clients and servers behave. However, application programming interface (API) is a more official term. On the other hand, API is a more comprehensive set of protocols that may be applied to many platforms to help it connect with other software. REST is compatible only with web applications. Most of the time, it manages HTTP requests and responses. Next, we show how a stateful REST API–based fuzzer can be expanded with an active property checker. These regulations automatically evaluate and detect violations. The first stateful web service is REST. You may use a REST API fuzzing tool to automatically test cloud services by using their REST APIs. APIs may also be used to find dependability and security holes in those services. We describe how to apply these checkers in a remarkably consistent and time-saving way, and we frequently use these checks to address new issues found in various Azure and Office365 cloud deployments. Finally, we discuss the potential security risks of these services.

O. Sri Nagesh, P. Laxmikanth, G. Raja Vikram, V. S. S. P. L. N. Balaji Lanka
Study and Research on Autism Spectrum Disorder Using Supervised Machine-Learning Techniques

Machine learning (ML) is embedded in everyday life and is used around the world by various organizations and in various fields. It has brought about a drastic difference in the healthcare industry. In addition, autism spectrum disorder (ASD) unknowingly affects people, but we are now predicting ASD in advance, on the basis of various behaviors, mainly the way that people interact/communicate with others, the way that they perform tasks, and their reactions. Detecting ASD at an early stage is essential to controlling it and preventing it from worsening. In this study, we incorporated child, adolescent, and adult datasets, and we used logistic regression, K-nearest number (KNN), random forest, support vector machine (SVM), and Naive Bayes to predict whether a person has ASD. To determine the accuracy of the above algorithm, we have used 20% data for testing and the remaining for training the algorithms. The outcomes of the various techniques are as follows: Logistic regression had an accuracy value of 95.0%, and KNN, random forest, SVM, and Naive Bayes had accuracy values of 84.72%, 96.36%, 86.36%, and 95.15%, respectively.

P. Vinod Babu, Kakarla Sanchit, Nimmadala Nithin, P. Saketh, Nekkanti M. Sri Rama Chowdary
Semi-Automated Vehicle Controlled Using Wi-Fi

A semi-automated vehicle, equipped with a camera, can detect lanes on the road and drive itself in the lane. And when the lane detection process cannot help the model in finding the lane, then the user can instruct the semi-automated vehicle using keyboard commands that are taken as inputs by Raspberry Pi, and the processed output will be passed to Arduino, which is connected to each other and also has a Wi-Fi (wireless fidelity) connection. This Wi-Fi connection helps in transmitting the user inputs to the Raspberry Pi. The Arduino, a slave device, manipulates the speed and direction of the motors, depending on the inputs from the master device (Raspberry pi).

D. K. V. Dhanush, Ch. Murari, K. Manjunath, G. Saketh, Mrudula Owk
Disease Identification System for Aura Images Using Fruit Fly Optimization (FAO) Technique

This chapter presents a novel methodology for disease identification acquired from the aura images of the individuals and endorsing the disease using the color image processing technique. The fruit fly optimization algorithm (FOA) is utilized for effectively identifying the feature vector using which the disease identification is carried out. The methodology is carried out using the Bio-Well benchmark data set and the results derived are subjected to evaluation based on qualitative metrics such as average difference (AD), minimum difference (MD), and image fidelity (IF). Image segmentation quality metrics are also considered, like the Probability Random Index (PRI), Global Consistency Error (GCE), and Volume of Information (VOI), to identify the accuracy of the segmentation procedure. The results derived showcase a good recognition accuracy of around 93%.

Manjula Poojary, Yarramalle Srinivas
Hybrid Movie Recommendation System Based on User Preferences and Item Similarity

In recent years, recommendation systems have gained popularity for providing users with relevant information to aid them in decision-making from vast amounts of available data. Movie recommendation systems rely on either user similarity (collaborative filtering) or specific user preferences (content-based filtering) to generate recommendations. Often, both methods are combined to enhance the effectiveness of recommendation systems, alongside various similarity measures to determine user likeness for recommendations. This study investigates current techniques, including content-based filtering, collaborative filtering, hybrid approach, and unsupervised and association rule mining algorithms for a movie recommendation, along with various similarity measures. These strategies enable recommendation systems to provide personalized recommendations to users and improve their experience.

Sundara Siva Rao Ivaturi, Mohana Naga Vamsi Thalatam, Aravind Siva Kumar Bugatha, Sruti Sudha Ladi, Dharma Teja Botta
Comparison of Artificial Intelligent Systems for Real-Time Accident-Prone Applications

Fatigue drowsiness is a major cause of road accidents, particularly among long-distance drivers. To address this issue, numerous artificial intelligence (AI) solutions for detecting driver fatigue in real time have been developed. We performed an analysis of four distinct AI algorithms for detecting driver tiredness in this study: 2 s-STGCN, PERCLOS (percent of the time eyelids are CLOSed), support vector machines (SVM), non-negative matrix factorization (NMF), convolutional neural networks (CNN), and AlexNet. We used publicly available datasets of more than 3000 images collected from drivers under different conditions, including varying levels of drowsiness. The dataset was preprocessed and augmented to improve the robustness of the models. We then trained and evaluated each of the four AI techniques on the dataset using various performance metrics, including accuracy. Our results showed that all four AI techniques performed reasonably well, with AlexNet achieving the highest accuracy of 99.65%. SVM and CNN with PERCLOS also performed well, with accuracy scores of 94% and 93.4%, respectively. CNN and 2 s-STGCN had lower accuracy scores of 93.37%. Our findings suggest that AlexNet is the most effective AI technique for driver drowsiness detection, followed by SVM and CNN. These findings can be used to help build robust driver drowsiness monitoring tools, which may also help minimize the number of car accidents that result from driver fatigue.

Venkata Subba Rao Are, Anuradha T., Pooja Nagabhairu, Geetha Sai Putty, Anudeep Peddi, Chandra Sekhara Rao M. V. P.
Hand Talk Assistance with TensorFlow Single Shot Detector

Hearing-impaired people use sign language as their primary form of communication worldwide. A verbally disabled person and a normal person have, nevertheless, always had trouble communicating. Real-time recognition of sign language is a huge improvement in their ability to communicate with one another. To communicate within their communities and with others, deaf, hard-of-hearing, and other people who are unable to speak orally use sign language. As a result of the prior system’s incapability to identify various indications, we are using it in this chapter to inform the deep learning-based method we are using to find a solution. In this chapter, we are going to apply the Single Shot Detector algorithm to detect different types of hand gestures and label them accordingly. So, whenever we capture an image through a webcam, our model will detect hand gestures, predict the type of sign, and display the label of that particular sign.

Dabbeeru Priyanka, Malla Moulya, Tangudu Sai Lakshmi, Gedela Satwika, Allu Aravind, Kagada Uday Kiran
A New Text Representation Technique-Based Approach for Authorship Verification

Author verification predicts whether the given text is written by the suspected author or not. Researchers used different types of stylistic features to differentiate the writing style of the author. PAN is an organization conducting competitions on different tasks every year by providing a suitable dataset. The author verification task has been included in several years of PAN competition. In this article, we conducted an experiment on the PAN competition 2022 author verification task. In this task, we need to predict whether the given two texts that belong to two different discourse types are written by the same author or different authors. The dataset contains an English corpus, which contains the pairs of texts written using four discourse types, such as business memos, text messages, emails and essays. We proposed a new text representation technique for the task of authorship verification. In this approach, we used word embedding techniques for representing words as vectors. The word embedding techniques consider the importance of a word in the total dataset of documents to generate the word vectors. The TFIDF measure is used to identify the importance of a word within a document. Each word is represented by combining the TFIDF weight of word and word embedding vector. The documents are represented as vectors by aggregating the word vectors that are contained in the document. These document vectors are trained with machine learning algorithms for predicting the accuracy of the proposed approach for authorship verification. We identified that the proposed text representation technique attained the best accuracies when compared with various solutions for authorship verification.

T. Raghunadha Reddy, P. Vijaya Pal Reddy
Early Recognition and Ranking of Knee Osteoarthritis by the Assistance of Enhanced Deep Learning on Knee MR Image Data

Osteoarthritis (OA) of the knee is an inflammation that impacts the knee bone due to the significant weight-bearing of the body. The disease results in degeneration and rupture of the cartilage elements in the knee joint, causing severe pain. Unfortunately, the prevalence of OA has been increasing globally, with a 113.25% increase in cases from 1990 to 2019. Currently, more than 350 million people globally suffer from arthritis, and it is estimated that by 2040, about 78 million US adults will have this condition. The diagnosis of OA is primarily carried out by evaluating symptoms and comparing plain radiographs, which can be subjective. However, Convolution Neural Networks (CNNs), one of the best deep learning technological advances, are currently attracting interest as a potential remedy for healthcare problems. Consequently, this study’s objective aims to design and implement a categorization scheme that can aid doctors in reducing their workload and assist rheumatologists in assessing the severity of the pain accurately. Furthermore, this will enable them to make the best diagnosis and recommend the most appropriate treatment. By using our proposed model, i.e., Enhanced MobileNet-V2, rheumatologists can make informed decisions on the severity of the condition, which can lead to better treatment outcomes for the patients. The proposed model Enhanced MobileNet-V2 was trained on the digital knee X-ray image dataset and achieved 91% accuracy.

Molleti Bala Murali, Varigala Sai Purnima, Kodamanchili Venkata Laxmi, Rajana Sai Sampath, Mohan Mahanty
Smart Blind Stick with Wristband: Obstacle Detection and Warning System

The smart blind stick-connected wristwatch is a device designed to assist visually impaired individuals with independent navigation while also providing additional features to enhance their daily life. The device combines the functionality of a traditional blind stick with the convenience of a wearable wristwatch, making it a discreet and convenient option for users. The device includes sensors and technologies such as object recognition and environmental sensors to provide feedback to the user, allowing them to navigate safely and confidently. The object recognition feature uses machine learning algorithms and computer vision to identify obstacles in the user’s path, such as stairs or curbs, and provides feedback through vibrations or audio cues. The environmental sensors provide information about the weather or other environmental conditions to help the user plan their route. The device also includes an emergency response feature that allows the user to quickly call for assistance if needed, providing peace of mind and a sense of security. The blind stick–connected wristwatch provides a comprehensive solution for visually impaired individuals, combining navigation, connectivity, and safety features in a discreet and convenient wearable device.

R. Ramyadevi, R. Loganathan, R. Karthikeyan, A. Vijay
Crop Recommendation System Using Machine Learning

Given that it accounts for 17% of the nation’s GDP and employs more than 60% of the workforce, agriculture is one of India’s largest and most diverse economic sectors. Numerous biotic along with abiotic parameters are used to make crop suggestions to boost agricultural productivity. It helps keep food prices down and favors farmers and the entire nation. Indian farmers frequently struggle with the issue of improper crop choice concerning the needs of their land. As a result, their output has been severely hindered. So, the crop recommendation model in this chapter thus uses research-based data on soil characteristics and soil classifications for farms to select the appropriate according to site-specific crop factors. By analyzing the datasets and applying machine-learning classifiers, including Decision Tree, Logistic Regression, and Random Forest, this model calculates the optimal crop for each soil type. As a result, choosing the right crop is easier, increasing productivity.

G. Gayatri, K. N. V. S. Praharsha, K. Hemanth, Mrudula Owk
Exploring Public Perception and Opinion Trends on Agnipath Scheme Through Sentiment Analysis and Topic Modeling of Tweets

Microblogging is a form of online writing that allows users to share short updates or media, such as pictures and audio recordings. A variety of microblogging platforms are available, such as Plurk, Tumblr, and Emote.in, among others. Twitter, however, is the most popular and rapidly growing platform, with millions of text posts. People worldwide can express their ideas using “tweets,” which may include text, images, videos, and audio clips. Twitter is a valuable source of content uploaded and shared by users, with around 204 plus million people posting an average of 411 million tweets per day. This content can be analyzed to obtain insight into consumer perceptions on various topics, including brand reputation, events, and political and sports discussions. In India, there are several ongoing social issues, such as Covid-19 booster shots and the Agnipath scheme. In this chapter, the authors analyze tweets using lexicon-based techniques to understand the Agnipath scheme’s global reach and gauge public sentiment.

Mohan Sai Dinesh Boddapati, Sri Aravind Desamsetti, S. Mahaboob Hussain, V. V. R. Maheswara Rao
A Feature Selection Technique–Based Approach for Author Profiling Using Word Embedding Techniques

Textual data are tremendously increasing on the World Wide Web through different social networking platforms, such as blogs, review sites, Facebook, Instagram, Twitter, etc. Crimes or frauds are also growing with the development of text in these platforms. Most of the users on these platforms create accounts with false information and send harassing and threatening messages. Knowing the details or basic information of text posted on these platforms is a very important task to control fraud. Author profiling (AP) is one technique used by the research community and social network administrators to know the author’s information related to messages posted on social networking platforms. Author profiling is a method of detecting demographic profiles such as age, gender, education, etc., of authors based on their written texts. Author profiling techniques are used in various applications of text processing, such as marketing, forensic analysis, and security-related fields. Most author profiling techniques use the content information of text to differentiate the writing styles of authors. The dataset contains a huge number of content words and, the identification of relevant words for differentiating the writing style of different authors is one challenging task to the research community. Feature selection algorithms are proposed by the researchers to find relevant features in the dataset. In this article, we developed a feature selection technique–based approach by using word embedding techniques for predicting the gender and age from the selected dataset. The PAN competition author profiling datasets are considered in this article for experimentation. In the proposed approach, word embedding techniques are used for converting words into word vectors. The similarity measure is used for finding the similarity among the word vectors. Similar words are grouped into clusters. Select the most important words from these clusters and consider them for experimentation. Selected words are used for representing the documents as vectors. These vectors are passed to machine learning (ML) algorithms to develop a model for classification. This model predicts the accuracy of the proposed method. The proposed approach attained the best accuracies for gender and age prediction when compared with the accuracies of several well-known approaches.

Karunakar Kavuri, M. Kavitha
Pothole Detection Using IoT to Help People

Potholes are known from a long ago, and their solutions are also found from different angles. While doing the literature survey, there were many researchers and solutions that ccould be applied with the help of different hardware systems. Detection of potholes is not the only thing that can be used to avoid them. Displaying potholes is also important, which can make drivers aware of them. The system consists of a GPS module that will be collecting coordinates, and an ultrasonic sensor will sense the distance, which will then be used to find the average distance from the road after every few cycles. Getting these two data, which will be coordinated and filtered out according to the limit suggested by the investigators of the UK (which is more than 40 mm), will be sent to the Cloud and can further be displayed on an Android App with the help of Google Maps. This system will not only help to detect potholes but also, by locating them, they can be avoided as well as soon repaired for future scope.

R. Suganya, Boya Harshita, Dhana Lakshmi, M. Mithun Kumar Reddy, Bada Rajasekhar Reddy
Machine Learning–Enhanced Diabetes Identification System

One of the most deadly chronic conditions that raise blood sugar is diabetes. Diabetes can cause a wide range of issues if it is not treated or recognized in time. A patient ultimately visits a diagnostic facility and sees a doctor there rather than going through the time-consuming identification procedure. On the other side, machine learning deals with this important problem. We have created a model that can successfully predict whether a patient will eventually acquire diabetes. Four machine learning classification techniques were utilized in this experiment to identify diabetes early: SVM, Gradient Boosting Classifier, KNN, and Naive Bayes. The studies make use of the PIDD (Pima Indians Diabetes Database) from UCI’s machine learning repository. These measures are used to assess all algorithms’ accuracy, precision, F-measure, and recall. Accuracy is assessed using instances that were properly and erroneously categorized. The findings show that accuracy is better for Support Vector Machines than for other methods. Receiver operating characteristic (ROC) curves are used to correctly and methodically confirm these findings. We are able to decrease the number of features that must be included or deleted in our prediction model, improving its accuracy, by combining Recursive Feature Elimination with hyperparameter adjustment.

Rama Rao Adimalla, Tirlingi Thirupathirao, Ravindranath Gatte, K. Satynarayana Murthy, Bosubabu Sambana
Diabetes Detection and Analysis Using Machine Learning

One of the most lethal chronic illnesses is diabetes, which leads to high blood sugar levels. Many complications might arise from diabetes if it is not properly diagnosed or treated. Seeing a doctor and going through the identification procedure takes too much time, therefore patients usually resort to diagnostic centers instead. However, this major problem can be solved with machine learning. In order to correctly forecast whether or not a patient would acquire diabetes, we have created a model. In this study, four different machine learning classification methods were utilized to diagnose diabetes at an early stage: SVM, Gradient Boosting Classifier, KNN, and Naive Bayes. UCI’s machine learning repository provides the PIDD (Pima Indians Diabetes Database) for use in the tests. These measures are used to assess the algorithms' accuracy, precision, F-measure, and recall. Accuracy is measured by comparing instances with the right and wrong labels. The findings show that Support Vector Machine is the most effective algorithm. These results are verified in a thorough and methodical manner by using ROC curves. Reduce the number of features that must be included or deleted in our prediction model by combining Recursive Feature Elimination with hyperparameter adjustment.

Bosubabu Sambana, P. Srinivasa Rao, Adimalla Rama Rao, Chinnam Yuvaraju, T. Murali Mohan, P. J V G Prakasa Rao, M. N P Patnaik, Dekka Satish
Identification and Analysis of Neural Disorders Based on Hyperactivity and Spectrum Disorder with Functional MRI

A collection of conditions known as neurodevelopmental disorders affect a number of factors, including mood, learning capacity, self-control, and memory. The most frequent and prominent are attention deficit hyperactivity disorder (ADHD) and autism spectrum disorder (ASD). The current diagnosis methods are time-consuming, human-mediated, and unreliable. This study focuses on the prediction of ADHD and ASD, and a comparative analysis is done using neural networks, support vector machine, gradient boosting, and logistic regression models to determine which performs well for the resting-state functional MRI (fMRI) data. During this procedure, time-series signals were collected from the brain region’s voxels and utilized to create functional connectivity characteristics. Our models were evaluated on various metrics, and we observed that a multi-layer perceptron model performs well.

Sai Siddharth Kodukula, Usha Bala Varanasi, Soumya Boddikura, Pavani Talluri, Bala Sai Kalepu
Review Analysis Using Web Scraping in Python

Analyzing customers’ problems to increase maximum sales on the online market. The description hashtags can be provided here. This work will eradicate the problem of sellers who want to keep their products at the top of any online web applications. This analysis aims to explore the numerous features and build a classification model that can tag the rating based on the description and consequently rate them as positive or negative. This analysis will focus on the score, summary, description and score-based sentiment features. The reviews were gathered from the cloud framework (AWS) and stored the reviews, which were newly assigned through the Webpage that has been developed. The goal was to build a model for text classification by pre-processed data using various techniques and libraries. Then create a word cloud plot based on the high and low scores summary. The pre-processed data is converted to numbers (vectorized) to feed the data into the model. Then train the model and optimize the parameters, which will increase the overall accuracy. After building the classification model, results will be predicted for the test data. The project was developed in Python but used one of the cloud instances like Amazon Web Services, SAS, Google Cloud and Microsoft Azure, to find the best Text Mining tool that fits the project. At last, Python was chosen as the best option and AWS, for instance.

Namala Rahul Chowdary, Rita Roy, Bojja Anurag, S. V. N. S. Dakshina Murthy, Bodana Boina Nagarjuna Yadav
A Novel Approach of Deep Neural Network Performance Analysis for Predicting Chronic Kidney Disease

Chronic Kidney Disease (CKD) enters slowly into the human body but lasts for a long period after getting affected by the disease. This chronic disease cannot be completely eradicated kidney disease is a deadly disease that shows its intense reactions in the ending stages. That is why CKD identification in the early stage is required to provide the necessary treatment for the prevention or cure of such a deadly disease. Early detection of chronic kidney disease helps the patients and doctors to take appropriate care and treatment measures to cut down further consequences and progress of the disease. By using machine learning techniques in the healthcare sector, the pathological data of the patient, disease can be classified and predicted. Therefore, the indispensable requirement of the best and accurate methods for recognizing chronic kidney disease is observed in the medical sector. In this chapter, various experiments are conducted and observed that Deep Neural Network (DNN) is proved to generate accurate results in CKD recognition with 99% accuracy.

Ranjith Kumar Rupani, Ramakrishna C., V. Adilakshmi, Sabavath Raju
A Stroke Complication Neural Network Model to Predict the Severity of Brain Stroke Using Family History

These days, due to technological advancements, the lifestyle of modern humans has changed from being active to sedentary. As computational power is rapidly increasing day by day, we can use concepts like machine learning and neural networks to estimate and monitor the health of people by just using stroke parameters along with family history. The findings in the body of literature are already in existence, which show that brain strokes can be classified without considering the significance of family history. In this work, ischemic, intracerebral, and subarachnoid hemorrhagic brain strokes are the main emphasis. A Stroke Complication Neural Network (SCNN) is proposed. This model works by taking the stroke classification results from machine learning models. Using this stroke complication graph, family history, and a multi-layered perceptron (MLP), an attempt is made to determine the severity of the stroke. The proposed model has obtained decent outcomes with an accuracy of 94.32%.

Puneeth Gangarapu, Nitish Sine, Vamsi Bandi
Automating Curriculum Vitae Recommendation Processes Through Machine Learning

Most businesses now use Internet-based recruiting portals as their main hiring method. Such platforms save time and expenses associated with hiring new employees, but they have problems with outdated information retrieval strategies like Boolean search approaches. With an emphasis on two properties, we describe a CV recommender system in this study. The first characteristic is the capacity to automatically process candidates’ CV papers and classify them into positions. The capacity to suggest abilities to a candidate that are not included in their CV but are probably present is the second quality. Both features are based on skill extraction from a textual CV document, both features. For candidate categorization, a precomputed spectral skill clustering is used, and different similarity-based techniques are used for skill suggestion. In this project, we will utilize machine learning techniques such as Ngram and WordCloud to find the talents that are absent from the resume and to offer replacements. The experimental results show the efficiency of the suggested strategies through both automatic experiments and an empirical investigation.

Praveen Kumar Karri, D. Jaya Kumari, P. Laxmi Kanth, P. Ramamohan Rao, K. Sowmya Sree
Application Controlling Using Hand Gestures Through Yolov5s

With the aid of Python, OpenCV, Yolov5s, and PyAutoGui, we attempt to use hand gestures for controlling an application. The need for unfettered interaction has made it difficult for conventional input devices like the keyboard and mouse to be used. In linguistics, gestures with the hand are a crucial part of body language (Nagalapuram GD, Roopashree S, Varshashree D, Dheeraj D, Nazareth DJ, Controlling media player with hand gestures using convolutional neural network. In: 2021 IEEE Mysore sub section international conference (MysuruCon). 978-1-6654-3888-9/21/$31.00 ©2021 IEEE, https://doi.org/10.1109/MYSURUCON52639.2021.9641567 , 2021). Using the hand as a device makes it simple to interface with humans and computers. It would be intriguing to engage with machines using hand gestures (Li G, Li D, Yang A, Real-time hand gesture detection based on Yolov5s. In: Proceedings of the 41st Chinese control conference, July 25–27, Hefei, China, 2022). Different applications, Various software programs, including Media Player, PowerPoint, PDF viewers, and software that accepts keyboard and mouse input, are controlled by hand gestures. Interaction is simple, practical, and requires no tedious additional equipment when gestures are used (Paliwal M, Sharma G, Nath D, Rathore A, Mishra H, Mondal S, A dynamic hand gesture recognition system for controlling VLC media player. In: 2013 international conference on advances in technology and engineering (ICATE). IEEE, 2013). A camera-based hand detection system that employs a live webcam stream to identify the gesture once it has been observed. This system gives the user access to essential keyboard functions with the aid of the PyAutoGui package. With this technology, we can increase the usability of computers for people with mobility issues or who cannot have physical contact with peripherals. A deep learning algorithm called Yolov5s is used for object detection. The precision and recall are near to 1, the mAP (0.5) value for hand gesture recognition is 0.995, and the mAP (0.5: 0.95) value is 0.978. The Yolov5 model may more effectively fulfill the demands of real-time gesture recognition and can serve as a crucial practical foundation and point of reference for advancing human–computer interaction technology in the future (Jalab HA, Omer HK, Human computer interface using hand gesture recognition based on neural network. In 2015 5th national symposium on information technology: towards new smart world (NSITNSW), 2015).

K. Rajendra Prasad, Chadaram Tarun Naga Sai, Guntuboina Bhaskar Ganesh, Baina Rohini Raghavi, Gunnam Sri Saketh
Chili Leaves Disease Identification Using Artificial Neural Network Algorithms

Deep learning (DL) and machine learning (ML) approaches have made it easier to recognize and identify objects in photos. Following in the footsteps of their success in other industries, neural networks have lately entered a few agricultural and farming applications. Plant disease detection software may help farmers manage their crops more effectively and increase yields. Utilizing images to diagnose plant disease in crops is a difficult task in and of itself. The use of specialist control measures requires both the detection and identification of species. In this chapter, a study of research initiatives that used Google Net, a kind of DL, to solve different plant disease detection difficulties was undertaken. One hundred eighty images of chili illness from the Kaggle public domain were given to the Google-Net convolutional neural network (CNN) in order to assess training accuracy. The final training accuracy ranges from 30% to 100% by taking into consideration various learning characteristics, including CNN optimizers SGDM, ADAM, and RMSProp. Max Epochs and giving Dropout probability, Strides, Dilation factor, and padding values as constants. The Google-Net CNN architecture achieves 98.61% accuracy in SGDM and ADAM with Max Epochs of 20, 30, and 40, respectively, according to the simulation. Moreover, it was claimed that SGDM is suitable for training the Chili Dataset, whose Max Epochs are substantially smaller than in SGDM and ADAM. The two optimizers have three epochs: 20, 30, and 40, with low epoch 20 producing superior accuracy.

Kanaparti Kantharaju
Correction to: Investigating Context-Aware Sentiment Classification Using Machine Learning Algorithms
P. Ashok Kumar, B. Vishnu Vardhan, Pandi Chiranjeevi
Backmatter
Metadata
Title
Accelerating Discoveries in Data Science and Artificial Intelligence I
Editors
Frank M. Lin
Ashokkumar Patel
Nishtha Kesswani
Bosubabu Sambana
Copyright Year
2024
Electronic ISBN
978-3-031-51167-7
Print ISBN
978-3-031-51166-0
DOI
https://doi.org/10.1007/978-3-031-51167-7

Premium Partner