Advanced Network Technologies and Intelligent Computing

The COVID-19 pandemic places additional constraints on hospitals and medical services. Understanding the period for support requirements for COVID-19 infected admitted to hospitals is critical for resource distribution planning in hospitals, particularly in resource-reserved settings. Machine Learning techniques are being used to approximate a patient’s duration of stay in the hospital. This research uses Decision Tree, Random Forest and K-Nearest Neighbors, Voting classifiers, and Stacking classifiers to predict patients’ length of stay in the hospital. Due to the imbalance in the dataset, Adaptive Synthetic (ADASYN) was used to resolve the issue, and the permutation feature importance method was employed to find the feature importance scores in identifying important features during the models’ development process. The proposed “ADASEML” has shown superior performance to the earlier works, with an accuracy of 80%, precision of 78%, and recall of 80%.

Detection of neurological disorders such as Alzheimer, Epilepsy, etc. through electroencephalogram (EEG) signal analysis has become increasingly popular in recent years. Alcoholism is one of the severe brain disorders that not only affects the nervous system but also leads to behavioural issues. This work presents a weighted visibility graph (WVG) approach for the detection of alcoholism, which consists of three phases. The first phase maps the EEG signals to WVG. Then, the second phase extracts important network features, viz., modularity, average weighted degree, weighted clustering coefficient, and average degree. It further identifies the most significant channels and combines their discriminative features to form feature vectors. Then, these feature vectors are trained by different machine learning classifiers in the third phase, achieving 98.91% classification accuracy. The visibility graph (VG) is not only robust to noise, but it also inherits many dynamical properties of EEG time series. Moreover, preserving weight on the links in VG aids in detecting sudden changes in the EEG signal. Experimental analysis of the alcoholic EEG signals indicates that the average accuracy of the proposed approach is higher or comparable to other reported studies.

In the process of production of Areca nut, segregation is one of the important stages. As of now, most commercial retailers use skilled workers for quality segregation, which means a lot of time is required for finalising the product costing. In this paper, Convolutional Neural Network (CNN) and MobileNet based methodology was proposed to identify the healthy and diseased Areca nut. The dataset containing images of healthy and diseased nuts was created. Furthermore, the augmentation method was applied to enhance the dataset. The confusion matrix was applied to check the performance of the models and the same was cross-validated using the 10-fold method. The CNN and MobileNet achieved accuracy, precision, recall, and an F1-score of 100%. After applying the 10-fold method, the CNN achieved average accuracy, precision, recall, and F1-score of 95%, 97%, 96%, and 94%, respectively. Whereas, MobileNet outperforms CNN by 100% for all metrics.

Customers should be one of the main focal points of any profitable business. Loyal customers who develop a relationship with the organization raise multitudes of business prospects. An organization looking to reap benefits from such opportunities must find a way to first of all, identify such customers and secondly, market their products to them in an individualized way to develop a lucrative business. This would require the organization to spot such customers and then differentiate their personal needs, preferences and behaviours. The aim of this paper is to tackle this problem using RFM analysis and Unsupervised Machine Learning technique called K-Means Clustering. RFM (Recency, Frequency, Monetary) analysis helps determine the behaviour of the customer with the organisation. The RFM values for each customer are calculated first following with the RFM Scores. Then, K-Means Clustering is implemented on the basis of the RFM Scores and in the end, we get clusters of customers. At this point, we will be able to analyze each cluster and accurately identify the characteristics of the customers. This will make it easy for the organization to customize their marketing strategies according to the customer behaviour, which will result in raised profits.

Several medical specialists today use X-ray and CT scan pictures of the lungs to classify respiratory disorders for specific diagnosis. Respiratory illness categorization based on inhaling and gasping sounds is still a work in progress in the scientific area. Respiratory illnesses have a high fatality rate among some of the chronic diseases. Early identification of respiratory disorders is critical for lowering death rates and curing illness. The automated categorization of respiratory sounds seems to have the capability to identify irregularities in the early phases of a respiratory disorder and, as a result, boost the efficacy of decision-making process. In this paper, a novel approach for classifying respiratory diseases using manifold D-CNN (Deep Convolutional Neural Network) is proposed to contrast different respiratory diseases accurately. Audio features are extracted using LibROSA (Mel-frequency Cepstral Coefficients, Spectrogram, and Chromagram). The performance of the model is tested using Respiratory Sound database and an accuracy of 96.23% is obtained, thus ensuring its supremacy over numerous state-of-art comparative models.

Cardiac diseases are diseases that affect people across the globe, and cardiac failure occurs without any warning. Identification of cardiac diseases at an early stage becomes a challenge for researchers in the health domain. Machine learning frameworks and algorithms are effectively used in the current medical field to predict and classify various diseases accurately. In this paper, we explore the traditional supervised machine learning techniques and algorithms and their cardiac disease classification accuracy. We further investigate the feature extraction technique Kernel Principal Component Analysis with a pipelined framework. The proposed framework overcomes the issue of overfitting and increases the prediction accuracy most effectively. Random Forest produced the most perfect result and Extreme Gradient Boost technique achieved an accuracy of 99.02%. Other Boosting classifiers, Gradient Boosting and Light Gradient Boosted Machine produced an accuracy of 94.16% and 98.38% respectively.

Air Quality Index is an index that measures air quality on a daily basis. Poor quality of air is very harmful to human health. The study of the Air Quality Index is very necessary to know the air quality of a particular city or country. It creates awareness amongst the citizens about the air quality of a particular city or country. It thus, compels the authority to bring certain remedial measures to control air pollution and develop a proper system that works for the betterment of the air quality of a particular city or country. The main objective is to analyze the air quality index of Delhi, to study the pattern and to study the major factors contributing to the higher Air Quality Index. In this literature, a sturdy and robust framework for AQI prediction is proposed, where first done Data Collection is done, then Data Pre-processing, after that KNN Imputation, Exploratory Data Analysis (EDA), Outliers Detection, Comparison of AQI distribution year wise, quarter wise, month wise, and week wise, Feature Selection are implemented, followed by Data Standardization, Data Splitting, then different Machine Learning (ML) regressors are applied on the data set, finally 10-fold Cross-validation is implemented. Here, Random Forest Regressor has the highest 10-fold Cross-validation score of 88.27% and this regressor is used for building the best predictive model.

Cassava is one of the main sources of carbohydrates in the world. However, the diagnosis of diseases in cassava crops is laborious, time-consuming and requires specialised personnel. In addition, very little research is available on images of cassava leaves taken with mobile phones and under field conditions. Therefore, the study designs deep learning models for the detection of diseases in cassava leaves from photos taken with mobile phones in the field. This study used a dataset of 21’397 images of cassava bacterial blight, cassava brown streak disease, cassava green mottle and cassava mosaic disease from a Kaggle competition. Twelve CNN models have been evaluated by applying transfer learning and data augmentation. Each of the models was trained with uniform samples and class-weighted samples. The results showed that the use of weighted samples reduced F1 score and accuracy in all cases. Furthermore, the DenseNet169 model was outstanding with an accuracy and F1 score of 74.77% and 0.59 respectively. Finally, the causes that hinder correct classification have been identified. The results reveal that it is still necessary to work on creating a balanced and refined database.

The rapid development of digital data in today’s world needs highly efficient and automatic N-lingual image document classification and information extraction systems. Several text-based image processing systems have been introduced to date for different applications. However, they are limited to character recognition, word extraction, profiling, and script and language recognition. The proposed document classification and information extraction system categorizes the bilingual English-Hindi images into mutually exclusive categories by using SVM and random forest classifiers at the character-level and document-level recognitions, respectively. It used two pseudo-thesaurus, first to store the pre-defined English and Romanized Hindi characters and second to store the keywords of the pre-defined categories. It discriminates between the languages by determining the absence or presence of the shirorekha for the English or Hindi word images, respectively, and applies the minimum distance computing method to obtain the constructed words from the mapped English and Romanized Hindi characters. This system achieved promising results in extracting the information and classifying the bi-lingual image documents. It illustrates a case study to explain the working of the system.

The Indian Premier League (IPL) is one of the world’s most well-known league competitions that takes place in India during the summer, with people from the cricket fraternity competing for the coveted silverware. Using the concepts of machine learning, this research intends to help viewers gain good insight into how a team is placed in a particular scenario in a match and whether they have a good chance of winning or not. It will also assist investors and franchisees in determining which team to invest in to maximize profits. Coaches, sports analysts, and technicians also gain game facts and ideas about opposing teams, which aids them in making decisions and changing plans as needed. In this study, we have used three different ML algorithms to predict the percentage chances of winning of the competing teams: Logistic Regression, Decision Tree, and Random Forest. For this purpose, two datasets have been used, one containing the match result data and the other containing ball-wise data of the IPL matches played between 2008 and 2022. Our results show that the Random Forest algorithm achieves the highest accuracy score when compared to the other two algorithms. Also, an interactive web application has been developed and hosted using Streamlit so that users can use it as a service from any convenient device.

The exponentially thriving vehicle population in India, accelerated by the country’s growing population and economic growth, puts an extensive burden on traffic management in the country’s major cities and towns. As a result, Automatic Vehicle licence Plate Recognition (AVLPR) is an interesting and crucial area of research in the Intelligent Transportation System (ITS), particularly because it may need to operate in real time. To address this challenge and concern, an efficient multi-style Indian vehicle licence plate recognition system is proposed. In this methodology, we consider three phases: Image pre−processing, licence plate object detection and character recognition. In image pre−processing, a fuzzy based image contrast enhancement algorithm is used to enhance the quality of the image. Then, for object detection, You Look Only Once Version 5 (YOLOv5) is used. Finally, the characters in the licence plate are recognized using Licence Plate Recognition Network (LPRNet), an end-to-end deep Convolutional Neural Network (CNN) with Connectionist Temporal Classification (CTC) Loss. The experiment result determines that the proposed technique is efficient and accomplishes better character recognition accuracy for English language, Tamil languages and multi-style licence plates.

Blood Pressure (BP) is created when the heart pumps the blood into blood vessels. If this pressure is more than 140/90 mmHg, it is diagnosed as high blood pressure (HBP). If HBP is not noticed and treated at an early level, it may lead to life-threatening issues. The design and development of machine learning models (MLM), to predict HBP in advance based on bio-psychological factors is gaining the attention of people. MLM is assisting medical doctors in diagnosing diseases more accurately, though MLM is exceptionally doing great in this domain, they are data-dependent. Conventional MLM is evaluated for the considered dataset. The major pitfall of such a model is a dependency on the dataset. If the same model is exposed to different datasets of the same type, the performance of the model may not be consistent. This paper proposed a heuristic-based dynamic data-drive, Age, Anger level (AA)-anxiety level, cholesterol level, obesity level (ACO) based hybrid MLM to predict HBP. The proposed model initially calculates the degree of dependency in terms of Pearson correlation between the attributes and class label attributes. The model is said to be hybrid as it uses the correlation-driven apriori-based fuzzy association rule miner to predict HBP. The proposed approach is data-centric and dynamic, as it calculates the Pearson correlation value for the given dataset at runtime and also assigns the priority value to the attribute at run time. The experimental setup is done on 1100 data records; the proposed model has got 91.168% accuracy, precision of 0.946, and recall of 0.933. The output of the model is a fuzzy inference engine consisting of the top 10 meta-heuristic-based fuzzy association rules, these rules can be used by a person as a knowledge base to manage and treat HBP.

Machine vision analysis of echocardiography images (echo) has vital recent advances. Echocardiography images are ultrasound scans that present the cardiac structure and function that becomes helpful in a significant measure of eight standard echo views, namely A2C, A3C, A4C, A5C, PLAX, PSAA, PSAP, PASM of the Cardiac cycle, and also identifies the disorders. In this research, we introduce a vision model for echo analysis with a deep convolutional neural network protected by the U-Net, trained to phase the echoes, and extract information of the right ventricle, left atrium, aorta, septum, and outer internal organ wall. The data includes image bundles; input to the CNN model predicts the cardiac structure by a softmax function into different categories, which becomes an input to a U-Net architecture that encodes and decodes the layers and foretells the functioning of the heart through segmentation. In summary, the research covers designed architecture that presents state-of-the-art for investigating echocardiography information with its benefits and drawbacks continued by future work.

Nowadays, social media platforms are thronged with social bots spreading misinformation. Twitter has become the hotspot for social bots. These bots are either automated or semi-automated, spreading misinformation purposefully or not purposefully is influencing society’s perspective on different aspects of life. This tremendous increase in social bots has aroused huge interest in researchers. In this paper, we have proposed a social bot detection model using Random Forest Classifier, we also used Extreme Gradient Boost Classifier, Artificial Neural Network, and Decision Tree Classifier on the top 8 attributes, which are staunch. The attribute is selected after analyzing the preprocessed data set taken from Kaggle which contains 37446 Twitter accounts having both human and bots. The overall accuracy of the proposed model is above 83%. The result demonstrated that the model is feasible for high-accuracy social bot detection.

The increase in abusive content on online social media platforms is impacting the social life of online users. Use of offensive and hate speech has been making social media toxic. Homophobia and transphobia constitute offensive comments against LGBT + community. It becomes imperative to detect and handle these comments, to timely flag or issue a warning to users indulging in such behaviour. However, automated detection of such content is a challenging task, more so in Dravidian languages which are identified as low resource languages. Motivated by this, the paper attempts to explore applicability of different deep learning models for classification of the social media comments in Malayalam and Tamil languages as homophobic, transphobic and non-anti-LGBT + content. The popularly used deep learning models-Convolutional Neural Network (CNN), Long Short Term Memory (LSTM) using GloVe embedding and transformer-based learning models (Multilingual BERT and IndicBERT) are applied to the classification problem. Results obtained show that IndicBERT outperforms the other implemented models, with obtained weighted average F1-score of 0.86 and 0.77 for Malayalam and Tamil, respectively. Therefore, the present work confirms higher performance of IndicBERT on the given task on selected Dravidian languages.

Recognizing disease in coffee leaves is an important aspect of providing a better quality of coffee across the world. Economies of many countries in the world depend upon the export of coffee and if we fail to recognize the disease in the coffee plant it will have a negative impact on them. The objective of this paper is to propose models for recognizing disease in coffee leaf plants. For achieving our objective we used various pre-trained models. We used transfer learning approach to identify coffee leaf detection. There were several models that achieved great results on both training and testing data. However, the best-achieving model was VGG19 due to less memory utilized and less time required for execution.

In order to determine ticket prices, airlines use dynamic pricing methods which are based on demand estimation models. Since airlines only have a limited number of seats to sell on each flight, they usually regulate the price of the seats according to the required demand. During periods of high demand, airlines usually increase the fares of traveling and thus the rate of filling the seats gets slowed down. Whenever the seats gets unsold as the demand goes down, the airlines usually reduce the traveling fares in order to attract the customers. Since unsold seats represent loss of revenue, it will be more advantageous to sell those seats for a price above the cost of service per passenger. A major objective of this study was to determine the factors responsible for airline ticket price fluctuations, and explore the relationships between them (basically, predict how airlines will price their tickets). Thus a model was built which predicts air ticket prices in the future and thus consumers will be able to make better purchasing decisions. Data was downloaded from an online website and data pre-processing and the exploratory data analysis of the data set was done. Later, three different machine learning algorithms i.e., Linear regression, Random Forest and XGBoost regressor were used in this study to predict the price of airline tickets in India. Hyperparameter tuning was also done in order to achieve the best and accurate results for prediction. XGBoost regressor performed best results by achieving highest accuracy with $$R^2$$ R 2 of about 84% and least RMSE of about 1807.59.

The Images generated by high-resolution Synthetic Aperture Radar (SAR) have vast areas of application as they can work better in adverse light and weather conditions. One such area of application is in the military systems. This study is an attempt to explore the suitability of current state-of-the-art models introduced in the domain of computer vision for SAR target classification (Moving and Stationary Target Acquisition and Recognition (MSTAR)). Since the application of any solution produced for military systems would be strategic and real-time, accuracy is often not the only criterion to measure its performance. Other important parameters like prediction time and input resiliency are equally important. The paper deals with these issues in the context of SAR images. Experimental results show that deep learning models can be suitably applied in the domain of SAR image classification with the desired performance levels.

Robotic Process Automation known as RPA aims to automate business processes by reproducing human interactions with the graphical user interface. The implementation of a technology such as Robotic Process Automation (RPA) allows all these routines to be executed by software “robots”. The objective of this work is to develop a structural management framework for the implementation and control of RPA projects, based on the PDCA cycle and the RPA life cycle. To achieve this objective, a bibliographical analysis was carried out using key terms related to the theme. Few works related to the theme were identified. An analysis of the works was carried out, verifying that none of the works addresses all phases of the PDCA cycle. However, what is new is a structural framework that covers all phases of the PDCA cycle and the RPA lifecycle. In addition, this framework presents the functions of each of the stages of the RPA life cycle, necessary for the implementation and control of RPA projects, and presents the external/internal structure of the organization chart of an RPA team, passing through the various levels of implementation of RPA, given the complexity of this technology. Finally, a proposed methodology was also presented in the framework to assist in the creation of RPA KPI's. In short, this framework stands out from the others for being quite complete and being able to have good proposals for managing the implementation and control of RPA projects, in teams that are at different levels of RPA implementation.

Cryptography is the technique of hiding and transmitting confidential information via the internet. There are various traditional cryptographic approaches have been presented by different researchers. Still, these techniques have several limitations, including huge computational times (during key generation, encryption, and decryption) and a high level of complexity (during permutation and substitutions). In the current scenario, Lightweight electronic devices (mobile, IoT, Smart Home devices) are becoming increasingly popular today. These devices create massive amounts of data, making it necessary to seek shrinkable and lightweight cryptographic techniques to ensure security. In this paper, we employ the property of an Involutory Function and a Pseudorandom Number Generator (PRNG) for encryption and decryption. The Involutory functions overcome the permutation and combinations encryption process and shrink the cryptographic operations using Functional Based Encryption. The XORed and Bit-Level Encryption processes are used to acquire the encrypted image. The encryption process is examined in terms of Execution Time. The encrypted image is evaluated using standard NIST statistical testing, Correlation coefficient, and Histogram analysis, demonstrating that the encrypted data has excellent statistical properties and security performance in cryptographic applications.

Amidst this COVID-19 pandemic, it is of utmost importance to wear facemasks and follow precautionary and preventive measures to decrease the further spread of this virus. In recent years Convolutional Neural networks (CNN) has impacted tremendously in various fields for classification and detection systems. In this paper we propose a facemask detection system using deep learning algorithms and a comparative study of various metrics for these deep learning algorithms. Algorithms like VGG, Resnet, Inception, Nasnet and Densenet, and its variations have been used. Using these deep learning models as a base and fine tuning the output layers of these models we construct an architecture for deep learning. Hyperparameter tuning and other methods like data augmentation have also helped in achieving better results. Various metrics like precision, recall, F1score, Average precision, accuracy and hamming loss has been evaluated for the models trained. An accuracy of 93.76% and average precision of 90.99% is achieved for the Denset201. Furthermore, we propose a standalone facemask detection system using the Raspberry Pi and a camera by fitting in the Nasnet mobile model into the detection system. Many applications for this system can be foreseen in places like hospitals, malls, restaurants and other places of public interest as an authentication or entry access criteria system.

Social media mainly provides an unparalleled chance to detect depression early in young adults. Depression is an illness that so often requires the self-reporting of symptoms. Social networking sites can provide an ample amount of data and information to train an efficient deep learning model. We aim to perform depression analysis on Twitter by analyzing its linguistic markers, making it plausible to create a deep learning model capable of providing an individual discernment into their mental health much earlier than the traditional approaches. We use two models to detect depressive users using their tweets on this conquest, a simple Recurrent Neural Network (RNN) and Long-Short Term Memory (LSTM). The LSTM model outperforms simple RNN by having a validation accuracy of 96.21 and a validation loss of 0.1077. Both models were trained on a single customized dataset, half of which was from sentiment140, and the other half was extracted from Twitter using the Twitter API.

Agriculture Industry has evolved tremendously over the past few years. Numerous obstacles have been raised in the agricultural fields including change in climate, pollution, lack of land and resource, etc. To overcome these hurdles and increase the crop productivity, agricultural practices need to adopt smarter technologies. Crop Yield prediction at an early stage is an significant task in precision farming. The yield of any crop depends on many factors including crop genotype, climatic conditions, soil properties, fertilizers used, etc. In this work, we propose a framework based on machine learning technique to predict the yield of corn in 46 districts of Uttar Pradesh, the largest Indian state in terms of population over a period of 37 years. We combine weather data, climatic data, soil data and corn yield data to help farmers to predict the annual production of corn in their district. We implement Linear Regression (LR), Decision Tree (DT) Regression, Random Forest (RF) Regression, and ensemble Bagging Extreme Gradient Boosting (XGBoost) model. Upon evaluation of all models and comparing them we observe that Bagging XGBoost Regression model outperforms all other models with the accuracy of 93.8% and RMSE= 9.1.

There are over 9000 bird and frog species in the globe. Some of the species are rare to find, and even when they are, predicting their behaviour is challenging. There is an efficient and simple technique to recognise these frog and bird species contingent on their traits to solve this challenge. Also, humans are better at recognising birds and frogs through sounds than they are at recognising them through photographs. As a result, employed various CNN models including CNN-Sequential, CNN-ResNet, CNN-EfficientNet, CNN-VGG19 and a hybrid model Convolution Neural Networks with Long Short-term Memory (CNN-LSTM). It is a powerful deep learning model that has shown to be effective in image processing.Compared to standard alone models, hybrid model produces better accuracy. A hybrid system for classifying bird and frog species is provided in this study, which employs the Rainforest Connection Species Audio Detection dataset from Kaggle repository for both training and testing. The classification of bird or frog species by using audio dataset after processing it and convert it into spectrogram images. Among all the deployed models CNN-LSTM system has been shown to achieve satisfactory results in practise by building this dataset and achieves accuracy of 92.47.

In humans, Lifestyle Disease (LSD) is caused by an improper way of life such as less physical activity, sleeplessness, unhealthy eating habits, liquor drinking, and smoking. LSD leads to gastric problems, indigestion of food, and prognosis to heart problems, Type II diabetes, and lung diseases. LSD treatment and medication lead to high expenditure for patients and country through LSD management and policies. Patients who suffer from LSD need lifelong treatment. The solution to reducing mortality due to Lifestyle Diseases is early detection and effective treatment. LSDs are low progressive in nature and need an effective and accurate early prediction method for effective treatment. The most prevalent LSD, based on World Health Organization (WHO) statistics, are heart disease and diabetes problems. This proposed model identifies the influencing attributes for contributing disease risk such as diabetes and heart attack and their associations using novel feature selection techniques such as Novel Majority Voting Ensembled Feature Selection (NMVEFS) for heart disease (HD) and diabetes. The influencing attributes are used to build a Clinical Decision Support System (CDSS) for LSD using a deep neural network, which helps physicians in identifying heart disease and diabetes at an early stage with 97.5% prediction accuracy and decreases treatment cost.

High quality and sufficiently numerous data are fundamental to developing any machine learning model. In the absence of a prior estimate on the optimal amount of data needed for modeling a specific system, data collection ends up either producing too little for effective training or too much of it causing waste of critical resources. Here we examine the issue on some publicly available low-dimensional data sets by developing models with progressively larger data subsets and monitoring their predictive performances, employing random forest as the classification model in each case. We try to provide an initial guess for optimum data size requirement for a considered feature set size using Random Forest Classifier. This sample size is also suggested for other machine learning (ML) models, subject to their trainability on the dataset. We also investigate how the data quality impacts its size requirement by introducing incremental noise to original class labels. We observe that optimal data size remained robust for up to 2% class label errors, suggesting that ML models are capable of picking up the most informative data instances as long as there are sufficient number of objects to learn.

Conversational modeling is an important task in natural language understanding and machine intelligence. It makes sense for natural language to become the primary way in which we interact with devices because that is how humans communicate with each other. Thus, the possibility of having conversations with machines would make our interaction much more smooth and human-like. The natural language techniques need to be evolved to match the level of power and sophistication that users expect from virtual assistants. Although previous approaches exist, they are often restricted to specific domains and require handcrafted rules. The obvious problem lies in their inability to answer questions for which the rules were not written. To overcome this problem, we build a generative model neural conversation system using a deep LSTM Sequence to Sequence model with an attention mechanism. Our main emphasis is to build a generative model chatbot in open domain which can have a meaningful conversation with humans. We consider Reddit conversation datasets to train the model and applied turing test on the proposed model. The proposed chatbot model is compared with Cleverbot and the results are presented.

The High-Resolution chest CT scan images help to diagnose lung related diseases accurately. In general, the more advanced hardware used in CT Scan machines, the more high resolution images will be generated. But it is a costlier approach. This limitation can be overcome with the post processing of the images generated from the CT machine. Even when the image is upscaled, the quality of the image should be retained. So, the process of reconstructing the High-Resolution images from the Low-Resolution images is known as Image Super-Resolution. The recent advancements in hardware and Super Resolution deep neural networks enabled reconstructing High-Resolution images in an efficient way. The objective quality metric Peak-Signal-to-Noise-Ratio evaluates the performance of a SR deep model. In this paper, proposed a pixel attention based deep neural network, MediSR for chest CT scan medical image Super-Resolution. The model is trained with two chest CT datasets and the experimental results showed an improvement of 1.78% and 18.23% for the 2 $$\times $$ × and 4 $$\times $$ × scale factors over the existing literature.

COVID-19 profoundly impacts human beings in various ways, i.e., psychological, socioeconomic, fear, social isolation, etc., augmenting the prevailing inequalities in mental health. The role of machine learning (ML) can be understood through its various potential applications in Stress Prediction in mental health. This literature survey uncovered various related articles, which were utilized to determine the essential structure for analysis. The gathered information helped in providing the new ideas and the concepts, which were incorporated with the support of literature and classified under broad themes based on mental health during the pandemic COVID-19. This study emphasized assessing various existing “Stress Prediction Support Systems” based on machine learning. This article also addresses the mental health issues that were emerged due to COVID-19 pandemic, further; also analysed the previously available stress prediction Machine Learning based models.

The reproducibility of Computer Aided Diagnosis (CAD) in detecting schizophrenia using neuroimaging modalities can provide early diagnosis of the disease. Schizophrenia is a psychiatric disorder that can lead to structural abnormalities in the brain, causing delusions and hallucinations. Neuroimaging modality such as a structural Magnetic Resonance Imaging (sMRI) technique can capture these structural abnormalities in the brain. Utilizing Machine Learning (ML) as a potential diagnostic tool in detecting classification biomarkers can aid clinical measures and cater to recognizing the factors underlying schizophrenia. This paper proposes an ML based model for the detection of schizophrenia on the structural MRI dataset of 146 subjects. We sought to classify schizophrenia and healthy control using five ML classifiers: Support Vector Machine, Logistic Regression, Decision Tree, k-Nearest Neighbor, and Random Forest. The raw structural MRI scans have been pre-processed using techniques such as image selection, image conversion, gray scaling of MRI images, and image flattening. Further, we have tested the performance of the model using hold-out cross-validation and stratified 10-fold cross-validation techniques. The results showed that the SVM achieved high accuracy when the dataset was validated using a stratified 10-fold cross-validation technique. On the other hand, k-Nearest Neighbor performed better when the hold-out validation method was used to evaluate the classifier.

In this paper, we have developed a hybrid model to predict fake news that includes clickbait detection as a parameter. The correlation between two different labels has been computed using a chi-square test. After establishing the correlation, the clickbait implementation was done on the heading of the dataset, and a fake news detection model has been executed on the content of the dataset. Then, the results of both the models were combined to generate a hybrid model through a regex equation. Our model is successful in enhancing the accuracy of the existing models by 1–2%.CCS Concepts • Artificial Intelligence • Natural Language Processing • Information Extraction.

The page that houses a malicious snippet that could misuse a user's computing resources, steal confidential data, or carry out other forms of assaults is known as a malicious host URL. They are generally distributed across the world wide web under various usage categories like spam, malware, phishing, etc. Although numerous methods or fixes (to identify URLs) have been developed in recent years, still cyberattacks continue to occur.This study contributes towards implementing three tiers of the system for detection and protection from harmful URLs. The first tier focuses on evaluating the performance of discriminative features in model creation. Discriminative features are derived from URL details and “Whois” webpage information that helps in improving detection performance with less latency and low computational complexity. The influence of feature variation on Parametric (neural network) and non-parametric classifier detection results are assessed to narrow down to the most prominent features to be adapted in the best model for the task of identifying URLs with multi-categorization. The study reveals that non-parametric ensemble models like Light GBM, XGBoost, and Random Forest performed well with a detection accuracy of over 95%, which facilitated building a real-time detection system and differentiating multiple attack types (such as Malware, Phishing, and spam).The second tier focuses on validation with a global database to know, if entered URL is reported as suspicious by various detection engines already. If not, it enables the user in updating the global database with URL details that are new and not reported yet. Finally, the two modules are integrated to create a web application using Streamlit that provides full system protection against malicious URLs.

Monitoring the Land Use/Land Cover (LULC) changes in this present era has become the most demanding task and it is very crucial for planning, proper resource management, regulating the expansion in the fringe areas in the existing cities, etc. Now the collection of reliable data has become easier as the hyperspectral images captured by various remote sensing satellites are readily available in different spatial and spectral resolutions. Processing this data according to the Region of Interest (ROI) in order to extract meaningful information is challenging. The most crucial part of this process is to identify various land covers very accurately. Only an automatic classification provides a feasible solution, as the manual process is tedious, expensive, and time-consuming. This paper compares two different image classification algorithms in classifying the covers, which can be utilized for land use and land cover changing pattern analysis in the Varanasi district of India. The experiments were carried out with the two most popular classification algorithms, namely: The Maximum Likelihood classifier and the Minimum Distance classifier. The overall accuracy and kappa co-Efficient values computed are 41.67 and 0.12 for the Minimum Distance Classifier and 82.43 and 0.78 for the Maximum Likelihood Classifier. It has been observed that the Maximum Likelihood Classifier outperforms the Minimum Distance Classifier.

Quantum mechanics with radically novel properties: such as superposition, entanglement, and the no-cloning theorem, has begun to open up Quantum Computation beyond the scope of Classical Computation. In recent years, quantum technology has seen tremendous leaps both in academic research and commercial exploration. Quantum computation has started to find applications in many more new domains every day, from disrupting modern cryptography to enabling unconventional security techniques, from quantum chemistry to physical simulations, and from quantum machine learning to financial optimization. This paper first provides the conceptual groundings of quantum computation and explores blind quantum computation, a one-of-its-kind sub-categorization of quantum computing active research. The use of Blind Quantum Computation in Quantum Cryptography is elaborated in detail. Finally concludes with highlights of the applicability of the subject. The paper is an easy-to-follow guide introducing the research trends and open challenges for the new researcher in follow-up on the field.

Behavior prediction of surrounding vehicles is a critical task. The main goal of this work is the implementation of Gaussian Mixture Model (GMM) and K-means to predict behavior of other vehicles moving on the road and theoretical analysis of their performance. All the vehicles are clustered in three different clusters (aggressive, moderate and conservative) depicting their driving styles. In order to achieve this goal, we implemented GMM and K-means with the help of keras in Python 3.6. The models are tested with NGSIM (Next Generation Simulation) data on the US-101 and I-80 dataset. The results of both the algorithms are visualized using scatter plots. Statistical properties of driving styles are derived using the statistical properties of records belonging to that cluster. The performance of both the algorithms is compared.

Deep learning (DL) models have been used extensively for various applications such as image recognition, virtual chatbots, healthcare, and object detection tasks. DL models are trained with huge data for better prediction ability. It is difficult to collect a lot of data. Thus, use of transfer learning with even lesser samples may yield better recognition rate. However, there exist various pre-trained models which are used for transferring knowledge of one domain to other domain. As, they are trained over specific domain, they need to be fine-tuned as per target domain to improve detection rate. Therefore, this paper proposes six different models by using VGG16, VGG19, Xception, InceptionV3, DenseNet201, and MobileNetV2 respectively. Considering, agriculture domain, monitoring growth of plant is crucial. It may helP_In identifying issues early in plant such as nutrient deficiencies, diseases, weed infections, and affected by pests or insects. Thus, the proposed models are evaluated over cucumber plant stage dataset. Findings show that proposed model using VGG16 (P_VGG16) attains maximum testing accuracy of 97.98%. It is also obtained that P_VGG16 improves accuracy rate by 2% as compared to VGG16. This work also shows the comparison of proposed models with their respective original state-of-the-art pre-trained models.

Today, Smart home technology is commonly used for remote observation and control of devices and systems, for example, heating and lighting for convenience, support, and energy saving. Smart home devices incorporate the Internet of Things (IoT) to help automate activities based on the homeowners’ preferences by working together to share the home members’ usage data. Many papers are published on home occupancy detection. Occupancy and presence can be used in the contextual smart home to accurately determine the presence of someone in buildings or houses and also able to predict various events and pre-emptive action combines inexpensive, non-intrusive sensors including CO2, temperature, sound, light, and movement with the aid of supervised learning methods like quadratic random forest and support vector machine (SVM). This method primarily focuses on reliably predicting the total occupants in an area with the help of a combination of heterogeneous sensor nodes and ML algorithms with the greatest 98.4% and the highest 0.953 F1 score. This paper primarily focuses on reliably determining the people in a room utilizing numerous sensor nodes that are heterogeneous in nature and machine learning algorithms, using various parameters such as CO2, temperature, light, sound, and motion with the help of supervised learning methods like Logistic regression, Naive Bayes, SVM Linear Kernel, KNN, Decision tree, Random Forest, SVM RBF Kernel, with the Maximum Accuracy of 99.62% and F1 score 0.996 The effectiveness of a scaled dimensional data set was further assessed using linear discriminant analysis i.e. LDA and principal component analysis i.e. PCA.

Medical data is over and over again analyzed for disease diagnosis and proper treatment. Medical dataset usually contain missing data it is also treated as error. These missing values possibly will clue to incorrect disease diagnosis result. Meanwhile the medical data collection is costly, time incontrollable and an essential on the way to collected beginning various issues. Therefore get better missing data is an alternative of re-collecting the medical data. In this paper a Prediction Model has been proposed for missing data imputation in medical data. An experiment includes various datasets to validate the model as well as to establish the importance of imputation. A new Method name called enhanced random forest regression predictor is proposed for missing data imputation on medical dataset. Method is validated using 3 datasets named wisconsin, dermatology and breast cancer. All the datasets are downloaded from UCI repository. Missing data are generated manually in the original data from 1% to 15%. The proposed Prediction model is predict the missing values based on enhanced random forest regression predictor and evaluates the model using various classifiers. Classification is assessment of normal and abnormal disease diagnostics and produce the result of this experiment is accuracy. Proposed predictor has been compared with two imputation method as KNN and mice forest. Missing prediction model is perform better compared with other methods. Evaluation is demonstrating the classification and gives accuracy which is compared with original dataset and the imputed dataset. Missing data problem is a serious problem in medical data and can guidance downstream disease analysis. A proposed enhanced missing prediction model for missing data imputation is an application of imputing the missing data and disease analysis using classification in better way.

The Internet of Things (IoT) has a wide range of applications that improve our quality of life with smart things that can connect without human intervention. However, IoT networks face new threats similar to Wireless Sensor Networks. Concerning high risks seriously impact the network topology, security, privacy, and energy levels. A routing protocol for devices with limited resources in the Internet of Things networks is the IPv6 Routing Protocol for Low-Power and Lossy Networks (RPL). The nodes can be subject to several attacks when it transmits packets between nodes. One of the most potent attacks against RPL protocol is the DODAG Information Solicitation (DIS) Flooding attack, which has a detrimental effect on the node's limited processing power and energy level. The impact of the Hello-Flood attack has been examined for several scenarios in this study. The Contiki-Cooja simulation environment is used to carry out the experiments.

The global system of interconnection covers an enormous scope of multiple data resources and feasible access mechanisms. It provides a direction for developing robust frameworks to interact and exchange knowledge in a diversified domain. But a major problem arises when a fast response to queries are raised by end users from different corners of the worldwide forum. Quora is one of the web based platform for receiving questions and suggesting answers accordingly. Many a times it is observed that the queries seems to be very irrelevant and creates chaos situation. Researchers are still struggling to equip new policies and mechanisms to deal with virulent and non-virulent queries. The cognition behind this study, aims at designing of a model to classify sincere and insincere questions using soft computing mechanisms. In this paper, seven algorithms out of which three ensemble techniques are employed on the Kaggle dataset.

Facial expression recognition (FER) is vital in pattern recognition, artificial intelligence, and computer vision. It has diverse applications, including operator fatigue detection, automated tutoring systems, music for mood, mental state identification, and security. Image data collection, feature engineering, and classification are vital stages of FER. A comprehensive critical review of benchmarking datasets and feature engineering techniques used for FER is presented in this paper. Further, this paper critically analyzes the various conventional learning and deep learning methods for FER. It provides a baseline to other researchers about future aspects with the pros and cons of techniques developed so far.

The Indian meteorological department provides only short term forecasts of weather, but long term forecasting is required for effective planning of agro production related activities. Generally, long term rainfall forecasting can be achieved by two approaches, traditional almanac forecasting (TAF), and scientific weather forecasting (SWF). TAF is based on observations and experience using a combination of meteorological and astronomical indicators, activities of animals, insects, and plants, and almanacs (also called as panchangams). SWF is based on past records of climate in a particular region using mathematical models. The main objective of this study is to recommend crops based on forecasted rainfall in Tamilnadu, India using TAF and SWF models. In this work, a hybrid non-linear autoregression Neural Network is used as SWF and saint Kaikkadar Rainfall Prediction Methodology is used as TAF. An empirical study is conducted to compare the performance of TAF and SWF models It is clear that the hybrid AF NARNN model outperformed the baseline AF NARNN model in terms of RMSE value. The hybrid AF NARNN model improved forecast accuracy by 4.0%, 3.3%, 5.4%, 5.1%, 6.9%, 0.3%, 5.3%, 2.6%, 1.0%, 6.0, and 6.0, respectively, with the exception of October. The prediction of TAF model is more than 65% similar to SWF. As a result, the TAF model can be utilised as a tool for long-term forecasting of activities related to agriculture.

Diabetic retinopathy is a major issue faced all over the world peoples that causes permanent blindness. With the onset of symptoms of diabetic retinopathy and the illness advances to an extreme level, it is difficult to recognize diabetic retinopathy at an earlier level. This paper presents the automatic detection of blood vessel segmentation based on U-net architecture. First, the retina blood vessels were segmented using a U-Net Architecture with the encoder/decoder module of multiple convolutional neural networks. For segmentation, binary conversion techniques are used. For the classification, deep learning models were proposed, namely ResNet50, Inception V3, VGG-16, and modified CNN. The final results are measured on a standard benchmark DRIVE dataset that contains 2865 retinal blood vessel images. For image classification, the proposed modified CNN performed better for DRIVE datasets with an accuracy score of 98%. Precision of 98%, Recall is 94.5% and F1-score is 95%. This paper evaluates the perceptional quality of segmented retinal images using SSIM. In this study pixel intensity was measured using RMSE, and PSNR to assess the quality of the retinal vessel segmented image.

This paper is concerned with K-independent total traveling salesperson problem (KITTSP) which is a variant of the famous traveling salesperson problem (TSP). KITTSP seeks K mutually independent Hamiltonian tours such that the total cost of these K tours is minimized. KITTSP is an $$\mathcal{N}\mathcal{P}$$ N P -hard problem since it is a generalisation of TSP. KITTSP is a recently introduced problem, and, so far no solution approach exists in the literature for this problem. We have proposed six constructive heuristics to solve KITTSP which are the first approaches for this problem. We have evaluated the performance of these heuristics on an extensive range of TSPLIB instances and presented a detailed comparative study of their performance.

The importance of monsoon rains cannot be looked over, as it has an impact on activities all year round, from agricultural to industrial. In the domains of water resource management and agriculture, accurate rainfall estimation and forecast is extremely useful in making crucial decisions. This study presents various deep learning approaches such as Multi-layer Perceptron, Convolutional Neutral Network, Long Short-Term Memory Networks, and Wide Deep Neural Networks to forecast the Indian summer monsoon rainfall (ISMR) (June–September) based on seasonal and monthly time scales. For modeling purposes, the ISMR time series data sets are divided into two categories: (1) training data (1871–1960) and (2) testing data (1961–2016). Statistical analyses reveal ISMR’s dynamic nature, which couldn’t be predicted accurately by statistical and mathematical models. Therefore, this study provides a comparative analysis that demonstrates the effectiveness of various algorithms to forecast ISMR. Moreover, it also weighs the result with established existing models.

Now-a-days, Object detection algorithms becomes more popular because of their significant contribution to the field of computer vision. Object detection algorithms are divided into two approaches i) region-based approach and ii) region-free approach. In this paper, we implemented YOLOv4 and YOLOv5 techniques of region-free approach due to their high detection speed and accuracy. The objective of this paper is to identify the relevant and non-relevant parts of surveillance videos as watching entire video footage is a time-consuming process. The use case for this research work is ATM surveillance footage, where the dataset is publicly not available, to train the network we developed our data set and then train and compare the proposed models. After experimental results, it is observed that YOLOv5 archives 84% accuracy and give better results than YOLOv4 which achieve 56% accuracy only.