Advanced Computing

The most frequently used, long-lasting, and well-preserved human body part in forensic and anthropological investigations is the tooth. The tooth is said to be a great way to determine the biological profile of unidentified remains. A piece of evidence in cases where the recovered dead bodies are mutilated and dismembered beyond recognition, such as bomb blasts, terrorist attacks, airplane crashes, and other mass disasters. Results could not be available for many days while forensic experts use their skills to manually determine each person’s age and gender. A fully automated method was developed to ascertain the age and gender of a person based on digital images of their teeth. The procedure of determining gender and age from images of individuals is done methodically since teeth are a strong and unique part of the human body that persists for a longer length of time and exhibits less sensitivity to change in its natural structure. Dental evidence is regarded as valuable for determining sex when other body parts are insufficient, critical to the body, or unavailable. The main techniques for determining sex from teeth are visual/clinical, microscopic, and sophisticated techniques. This review article covered the difficulties and approaches like GNN which is used for sex determination from teeth.

Drowsiness and fatigue are significant contributors to road accidents. We can prevent them by ensuring adequate sleep before driving, consuming caffeine, or taking rest breaks when drowsiness symptoms appear Current methods for detecting drowsiness, such as EEG and ECG, are accurate but require contact measurement and have limitations for real-time monitoring while driving. Proposes using eye closing rate and yawning as indicators for detecting drowsiness in drivers, as a non-invasive and comfortable alternative the goal of this paper is to create a non-invasive system that can detect fatigue in humans and provide timely warnings. Long distance drivers who tend to not take breaks in between are always at a high risk of drowsiness. The primary behavioral indicators used in the suggested technique are the driver’s yawning and eye blinking. The purpose of this Problem is to alert the driver by detecting yawning via closed eyes or an opened mouth.

One of the simplest ways to tell someone else apart from you is by their face. A personal identification system like face recognition may use an individual’s traits to identify them. Detection of any Face and Stage are the two stages of process of the face recognition of human, which is used for facial image recognition modal (face recognition) in biometric technology. The Eigen face method and the Fisher face method are the two categories of methods that are frequently used in created facial recognition patterns. Principal Component Analysis (PCA) for countenance is used to reduce the number of faces in three-dimensional space by the Eigen face approach for image facial recognition. Finding the eigenvector that resembled the most crucial Eigen value of the face image was the major goal of applying PCA [1] on face recognition using Eigen faces [2]. Image processing is used in face detection systems with face recognition. This requires mat lab software, which is the required program. Neural networks are categorized as deep learning. Deep learning’s foundational component, feature learning, aims to obtain hierarchical information using hierarchical networks in order to address significant issues that previously required artificial design features. The framework used is termed as Deep Learning and it may include n number of significant algorithms.

This work presented analysis and implementation of driver drowsiness, distraction, and detection systems using image processing techniques. The literature review based on drowsiness, distraction, and detection have been taken with their parameters in tabulation form. Flow charts of software and hardware have been presented for the proposed architecture. A comparative analysis of parameters with their percentage of accuracy is given in the table. Therefore, the proposed system found better accuracy as compared to other results. After practically implementation, this system gives the accurate results for the detection of sleepiness of driver. It detects the driver’s state such as Sleepy, Drowsy & Active. The proposed work found accuracy in term of parameter like: eye detection accuracy is 95% and drowsiness accuracy is 90%, it is approximately around 5–7% more accurate as compare to other existing work.

Detection of an object and depth estimation is very crucial in the field of computer vision, facilitating tasks in the field of autonomous navigation, scene understanding and many more. There are lot challenges in the current existing technique such as occlusion and accuracy issues, impeding their real-world applicability. To surmount these limitations, the proposed work introduces an innovative approach that melds deep learning architectures with efficient computational methods. By fusing advanced object detection models with a sophisticated depth estimation network, the work proposed have achieved substantial enhancements in accuracy and precision. The proposed model pushes the envelope for real-time implementation, contributing to the advancement of object detection and depth estimation capabilities. This approach was augmented with a novel depth estimation technique, extracting diagonal pixel lengths and combining them with actual depths from the dataset. Subsequent analysis employed both linear and polynomial regression, revealing that the polynomial model (98% average accuracy) surpassed the linear model (80.96% accuracy). These findings highlighted the importance of capturing complex non-linear relationships between pixel length and object depth, showcasing YOLOv4’s robust object detection capabilities and emphasizing the significance of intricate depth estimation in visual cues.

The growing worldwide population and rapid technological breakthroughs have increased energy consumption, highlighting the need for renewable and eco-friendly energy sources. Biofuel uptake is difficult owing to high prices, requires significant government measures to compete with conventional fuels and biomass-to-biofuel conversion inefficiencies are problematic. This research shows how biofuels can alter sustainability and examine Gujarat’s biomass supply chain utilizing advanced forecasting and supply chain optimization methods. A dataset including 2148 unique locations spanning the years 2010 to 2017 was utilized, and afterwards subjected to clustering analysis resulting in the identification of eight different groups. The next two-year biomass production is projected utilizing AutoML techniques. Finally, the supply chain is optimized using Mixed Integer Linear Programming (MILP) in order to reduce both costs and carbon footprint, in accordance with the predicted value.

Qualities of coatings deposited by High-velocity oxy-fuel (HVOF) spray technique are sometimes greatly influenced by the deposition parameters. It is difficult to research and develop a comprehensive model of the HVOF spray process because of the complex chemical and thermodynamic processes involved. The aim of this study is to use a back propagation neural network to create a predictive model for the mechanical properties of NiCrSiBFe coatings deposited by HVOF. The impact of the deposition parameters with respect to the intermediate process is also examined in this study. The change in porosity, nano-hardness, and sliding wear rate of coatings under various powder feed rate, stand-off distance, and oxygen gas flow rate were predicted using back propagation neural network algorithm. Similar trends are seen when comparing the predicted and experimental results, indicating that the developed model correctly predicted the properties of NiCrSiBFe coatings. The average errors for porosity, nano-hardness, and sliding wear rate are 1.816%, 1.997%, and 4.405%, respectively. The developed back propagation model can therefore be applied to coating operating practice for spray performance prediction, and also for parameter management and optimisation.

The article aims to propose an analysis model based on the GAN (General Adversarial Network) algorithm to address the impact of uncertain factors on cost control decisions in power transmission and transformation projects (PTTP). This article deeply analyzes the uncertainty factors of power transmission and transformation engineering (PTTE), identifies the key factors that affect cost control, and uses GAN algorithm to simulate and predict them, improving the accuracy and reliability of the decision-making process. The research results indicate that the uncertainty cost control decision analysis model based on GAN algorithm can effectively improve the cost prediction accuracy of PTTP, with a maximum of 96.5%. This provides an important reference basis for engineering management and decision-making. Therefore, the article provides a new idea and method for cost control of PTTP, which has important theoretical and practical significance.

Power transmission and transformation project (PTTP) is a crucial part of the power system, and during the construction process, various special situations may be faced, such as adverse weather conditions, resource scarcity, etc. These factors may affect the project schedule. Therefore, the optimization of the construction period for PTTPs has important practical significance. The article conducted research on the optimization of construction period in special construction scenarios of PTTPs, and proposed a construction period optimization model based on BP (Back Propagation) neural network. Firstly, the special scenarios in the construction of PTTPs and the importance of schedule optimization were analyzed. Then, a schedule optimization model based on BP neural network was proposed, and the model was described and analyzed in detail. Subsequently, model validation and experimental analysis were conducted using actual case data, and the results showed that the model had good performance in optimizing the construction period, with a maximum optimization period of 3.7 days, while also improving safety and resource utilization.

Technology’s growing use has facilitated the quality of living. Artificial Intelligence (AI) is the field that aims to define how human intelligence is mimicked by machines which are programmed to think or behave like humans. Modern approaches and tools for evaluating human behavior have been made possible by modern advancements in the fields of machine learning (ML) and artificial intelligence (AI). Due to its applicability in several industries, comprising of entertainment, security and surveillance, health, and intelligent environments, human activity recognition has gained prominence significantly. Human activity recognition (HAR) using video sensors typically involves analyzing the visual data captured by cameras to classify and identify the actions of individuals. In the following paper, we propose ConvLSTM and LRCN-based Human Action Recognition. A huge variety of films from the publicly accessible data set, UCF50 comprises a wide range of activity classes that are used to build a statistical model. For the model proposed in this paper, the accuracy has turned out to be 94%, the average f1-score is 0.93 and the average recall is calculated to be 0.925. The Loss curve has also been plotted along with the accuracy curve for the proposed model for recognizing human activities.

This paper presents an innovative machine learning-based approach for predicting concrete rupture strength, offering a faster and more cost-effective alternative to traditional testing methods. The proposed methodology employs a Random Forest Regressor (RFR) model, surpassing other regression models like Decision Tree Regressor (DTR) and Linear Regression (LR). A user-friendly web interface has been developed to facilitate practical implementation. In addition to highlighting the cutting-edge solution for predicting concrete rupture strength, the paper outlines avenues for future research, including dataset expansion, advanced model exploration, real-time monitoring through IoT, environmental considerations, and industry collaboration for deployment.

HVAC (Heating Ventilation and Air-conditioning) play a vital role in various sectors, from residential and commercial to industrial applications. Understanding and optimizing the power consumption of these systems is crucial for energy efficiency and cost savings. This research aims to explore the power consumption of refrigeration systems during power ON mode and perform Exploratory Data Analysis (EDA) and Machine Learning (ML) algorithms to gain insights into factors influencing power consumption. The experimentation is conducted on refrigeration test rig and performance is calculated during power ON mode by adding NPCM (Nano-Phase Change Material) in evaporator section and comparison is to be made without implementation of NPCM in evaporator section. By utilizing ML algorithms, it becomes possible to create predictive models that can assist in optimizing the power consumption of refrigeration systems, reducing energy costs, and minimizing environmental impact. The accuracy of model by linear regression is around 66% by implementation NPCM in refrigeration system where as 23% model accuracy is found without implementation of NPCM in refrigeration system. Also it is observed that coefficient of performance of refrigeration system increase by around 15 to 18% as compared with without use of NPCM. Also power consumption is reduces to 5 to 7% with implementation of nano phase change material in refrigeration system.

In the current research work, a single cylinder spark (S.I)ignition engine were used for investigations of premium level gasoline-ethanol-alkane experimentally with different operating conditions e.g. variation spark ignition timing. The Engine Lab and PE3 software were used for engine control and data acquisition system. The data obtained after experimentation were used to predict the engine emissions for different operating conditions. The engine emission characteristics were predicted using three machine learning algorithmsviz linear regression, decision tree and random forest. It was found that emissions characteristics such as carbon monoxide, unburnt hydrocarbon found to be minimum for 24°bTDC experimentally as well as predicted by machine learning algorithms with different operating conditions than other spark timing positions such as 15°, 18°, 21°, 27°, 30° bTDC. All three machine learning algorithms gave better results but the random forest algorithm were more accurate than linear regression and decision trees.

Depression negatively affects the daily life of an individual and may even lead to suicidal tendencies. The problem is compounded by the scarcity of trained psychologists and psychiatrists in developing countries due to which many cases go undetected. The automated diagnosis of depression can, therefore, assist clinicians to screen the patients and help them to handle the symptoms. The advent of wearable devices in the past decade has helped in capturing signals, which can be used to diagnose depression. This work uses a publicly available dataset and develops a model based on the distribution of microstructures from the temporal data to accomplish the given task. The results are encouraging and better than the state-of-the-art. An accuracy of 86.90% is obtained by using the proposed pipeline. This work is part of a larger project that aims to detect depression using multi-modality data.

The “ELECTRA” program resolves important obstacles facing India EV field using adaptable MERN suite and mixing customer-facing and supporting functions. Some of the key features like Google Maps API in emergency braking, BBD100K dataset for assessing autonomous driving risks amongst others, have been quite significant contributors towards the achievement of the project’s objectives. This will involve front end development with interface that is easy to understand and has a journey calculator for those intending to use EVs. The backend builds strong server logic, important APIs, and a powerful database that uses Google Maps API to provide timely information on EV charging stations and also improve journey planning. The use of the bdd100k dataset allows assessments in terms risk of emergency braking, which are crucial for the projects safety. Iterative testing process, user feedback, and adjustments improve platform performance and ease of use. Scalable, responsive and user friendly by deploying in production. Ease of use is promoted using user training and in-depth documentation. In terms of future development, upcoming periodic maintenance and constant improvement, mark the significance of the platform for Indian changing EV environment. ELECTRA helps in promoting the adoption of the electric vehicles through addressing issues like localizing charging points, and optimizing trip planning. Important milestones were achieved such as Journey Cost Calculator, integration of charger locations with google maps API, and utilization of YOLO in the emergency brake system.

This paper explores the domain of story generation and presents a novel approach that uses Generative Adversarial Networks (GANs), Recurrent Neural Networks (RNNs), and Long Short-Term Memory (LSTM) networks. The objective is to generate realistic and engaging stories for children. The traditional language models are proficient in maintaining grammatical consistency but often fail to establish long-term coherence. This study addresses and explores the individual performances and capabilities of 3 distinct text generation models based on GAN, RNN & LSTM respectively. The project employed three individual models and trained them on the same dataset and the evaluation was conducted using METEOR scores, accuracy, and loss metrics. To address this, our study introduces the use of GANs to enhance the quality of synthetic text. The MaskGAN model gave the highest accuracy and decent output on the trained dataset followed by the RNN and LSTM models. This paper is a significant step forward for story generation, highlights the unique contributions of GANs, RNNs, and LSTMs, amplifies the consistency and quality of independently generated narratives, and provides a foundation for future comparative analyses. MaskGAN achieved the highest accuracy and excelled in generating realistic and high-quality narratives followed by RNNs which exhibited decent accuracy but faced challenges with longer narratives, while the beam search-enhanced LSTM improved narrative quality offering a promising solution for coherent story generation.

Twitter is a micro-blogging website, which has amassed immense popularity over the past years. Many political parties are now using Twitter for publicity and running campaigns. These campaigns are run on various social media platforms to gain the attention of the voters. In this work, we analyze the effectiveness of such campaigns, by studying the sentiments that the users have towards the party and predict the result of the elections with its help. In this study, we utilise Hindi tweets for analyzing the sentiments that people have towards popular political parties in India. Various models were implemented and their performance was compared. The highest accuracy achieved was 88.4%.

Speech Recognition technology is widely used for voice-enabled form filling. The manual process of filling out forms by typing has become increasingly challenging and time-consuming. This issue is particularly evident in various locations such as job applications and internships. To address this problem, a solution is proposed as a system that automates the form-filling process using speech recognition technology. The ability to operate anything with voice command is a crucial factor in today’s environment. The proposed system is that it automatically fills out the forms. i.e., the system analyses the user’s unique voice, identifies the user’s speech, and then transcribes the speech into text. This paper proposes a machine-learning model that builds on Hidden Markov Model. The model will be trained and tested on this system and the proposed pre-processed methodology is Mel Frequency Cepstral Coefficients. The methodology was widely used in the prospect of recognition of voice automatically. The results demonstrate that this system effectively accurately transcribes user speech into text, simplifying the form-filling process significantly. By providing these results, we hope to demonstrate how this technology has the potential to revolutionize data entry and accessibility while also establishing a strong case for speech recognition as a convenient way to speed up form completion.

The use of information and communication technology for educational purposes has increased recently, and the development of network technologies has had a significant influence on the methods employed in electronic learning (E-learning). Most popular trends in education are e-learning, which shows how teaching and learning approaches are evolving in tandem with technological advancements. Technologies that facilitate education's scalability, automation, customization, and innovation offer enormous promise. The cost of e-learning has greatly decreased, and the benefits of its rapid, inexpensive, and time-saving instruction are substantial. Education technology (Edtech) solution providers helps to e-learning and guaranteeing that each student has a smooth and customized learning experience implies a significant impact on learners all across the world during the past COVID-19 epidemic. Thus, e-learning becomes popular teaching method in many educational institutions. However, online learning programs demand a physical examination by a real professor. Therefore, an automated evaluation system for learning prototype utilizing an excellent e-learning system is suggested in the current study. The Baye’s Theorem-based Bayesian Network (BN) concept can be best fit to construct intelligent e-learning systems. Groups of questions serve as the nodes and directed arcs serve as the edges of the directed acyclic graph (DAG) known as BN. This network is employed for ambiguous reasoning. The Baye’s network employing K2, KNN, and J48 was used to compare the BN model against AI classification techniques. Hence discovered that the performance of suggested smart e-learning approach utilizing BN outperforms that of the other two approaches, J48 and K-nearest.

This paper seeks to analyze and predict the course of Mastercard stock using three different Python libraries: SciKit Learn, XGBoost, and TensorFlow. This paper details information regarding machine learning algorithms and the linear regression model in particular. The paper presents the results of looking through the data and comparing some companies’ results with one another. Our study showed that leaner regression results with Scikit, XGBoost and TensorFlow library provide very high accuracy. The confident prediction for lower values, not to say the small increase in deviation for higher values was any worse.

This paper offers a comparative examination of two cuttingedge large language models, Guanaco and Llama, within the realm of natural language comprehension and generation tasks. Guanaco is a model fine-tuned on the open-source LLM Llama itself using Qlora, while Llama is trained on a combination of proprietary and open-source datasets. The assessment encompasses their performance on benchmarks like Massively Multitask Language Understanding (MMLU), Vicuna, and ARC. MMLU benchmark is a comprehensive evaluation of large language models’ capabilities on a wide range of tasks, including summarization, question answering, and natural language inference. ELO rating is a dynamic rating system that calculates the relative skill levels of players in zero-sum games, taking into account the outcome of each game. The abstraction and reasoning corpus (ARC) LLM benchmark is a set of tasks that are designed to evaluate the ability of large language models (LLMs) to reason and solve problems using only their core knowledge. The tasks are based on simple abstract concepts, such as objects, goal states, counting, and basic geometry. It demonstrates that Guanaco achieves strong performance on the MMLU benchmark, even outperforming Llama on the ARC benchmark. On the other hand, Llama excels on the Vicuna benchmark, surpassing Guanaco fine-tuned on open-source data. In a qualitative analysis, both models exhibit strengths and weaknesses. Guanaco showcases the ability to demonstrate theory of mind capabilities, whereas Llama sometimes generates inaccurate or unreliable responses in specific scenarios. Overall, this study sheds light on the performance and attributes of Guanaco and Llama, emphasizing their potential in various language comprehension and generation tasks.

The rise of AI-generated data, mainly from models like ChatGPT, LLAMA2 poses serious difficulties to academic integrity and raises worries about plagiarism. The current research looks on the competences of various AI content recognition algorithms to distinguish between human and AI-authored material. This research looks at numerous research papers, publication years, datasets, machine learning approaches, and the benefits and drawbacks of detection methods in AI text detection. Various datasets and machine learning techniques are employed, with various types of classifier emerging as a top performer. This work creates an Extra tree classifier that can distinguish ChatGPT produced text from human authored content. “ChatGPT Paraphrase” dataset was used for model training and testing. The result shows that the proposed model resulted in 80.1% accuracy and outperformed the existing models namely Linear Regression (LR), Support Vector Machine (SVM), Decision Tree, (DT), K-Nearest Neighbour (KNN), Ada Boost Classifier (ABC), Random Forest Classifier (RFC), Bagging Classifier (BG), Gradient Boosting Classifier (GBC).

Social media has changed the way people communicate, but it has also become a breeding ground for dangerous content. Natural Language Processing (NLP) is used in this study to classify unstructured data into dangerous and benign categories, providing insights about internet toxicity. The NLP approach used in the study gives light on the challenges and opportunities of toxicity identification. The researchers uncovered patterns and trends indicative of dangerous content by analysing massive amounts of text data, allowing them to construct powerful classification systems. The paper discusses the advantages and disadvantages of toxicity detection. Automated systems can swiftly scan enormous amounts of content, but they may misclassify some material, thereby leading to censorship or harassment. The online toxicity detection provide valuable guidance for stakeholders seeking to address this issue. By understanding the strengths and limitations of NLP-based approaches, informed decisions can be made about implementing effective toxicity detection strategies, ensuring a safer and more inclusive digital environment.

The objective of this study is to create a Handwritten English Alphabet Recognition System, emphasizing signature recognition. In a global context where handwritten records and signatures play pivotal roles in various sectors, including legal, finance, and authentication, the demand for accurate and efficient recognition methods is paramount. This research project endeavors to construct a resilient system that can precisely identify and categorize handwritten English letters and signatures through the application of machine learning techniques, notably deep learning. The system employs convolutional and recurrent neural networks to adapt to diverse writing styles and varying levels of complexity.

The stock price of a commodity is an essential factor for determining market volatility. Exact prediction of stock price and forecasting the market variation are crucial parameters of a stock analyst. The existing conventional approaches are incompetent to predict the stock market variations since they don’t take a comprehensive view but rather look at time-series data for every single stock. In this article, a time series relational model (TSRM) is proposed to predict the stock price. The proposed work combines the relationship between market conditions and price variation of a commodity with time. To anticipate stock prices, relationship information is collected using a graph convolutional network (GCN) and long short-term memory (LSTM) is used to extract time series information. This study attempts to forecast stock prices using the Time series technique, which is appropriate for the financial sector since stock prices fluctuate over time and involve the observation of varied changes regarding any given variable in regard to the respective time.

There has been a significant increase in recent years in the investigation of emotions expressed via speech signals; this field is known as Speech Emotion Recognition (SER). SER holds immense potential across various applications and serves as a pivotal bridge in enhancing Human-Computer Interaction. However, prevailing challenges such as diminished model accuracy in noisy environments have posed substantial obstacles in this field. To address the scarcity of robust data for SER, we adopted data augmentation techniques, encompassing noise injection, stretching, and pitch modification. Distinguishing our approach from recent literature, we harnessed multiple audio features, including Mel-Frequency Cepstral Coefficients (MFCCs), mel spectrograms, zero crossing rate, root mean square, and chroma. This paper employs Convolutional Neural Networks (CNNs) as the foundation for emotion classification. The Toronto Emotional Speech Set (TESS) and the Ryerson Audio-Visual Data-base of Emotional Speech and Song (RAVDESS) are two well-established datasets that we utilize. The accuracy of our proposed model on the RAVDESS dataset is 72%, and on the TESS dataset, it achieves an impressive 96.62%. These results surpass those of extant models that have been customized for each specific dataset.

The paper discusses how Large Language Models (LLMs) can be used in search engine optimization activities dedicated to e-commerce. In the first part the most important Search Engine Optimization (SEO) issues are discussed, such as technical SEO aspects, keyword selection, and content optimization. Then the study presents an in-depth look at OpenAI’s advancements, including ChatGPT and DALL-E. The latter sections describe the capabilities of Large Language Models into the realm of SEO, particularly in e-commerce. Firstly, a set of prompts for LLMs that can be used to create content and HTML code for online shops is proposed. Then advantages, and drawbacks of incorporating LLMs in SEO for e-commerce are presented. The research concludes by synthesizing the potential of merging AI with SEO practices, offering insights for future applications.

Handwriting is something which changes from person to person. Finding two people with same handwriting isn’t an easy job and not everyone can recognize all kinds of writing. But, in the growing era of technology and the modern world with the introduction of the domains like OpenCV – image processing and recognition isn’t a tough job. Further, with the growing dependency on technology and the ease of access, students can now solve equations at the comfort of their home. The job is simple, one just has to click picture of a problem written on the page, scan it, and the algorithm does it job. The system can recognize various handwritings and works on a large dataset. This Handwritten Equation Solver system, will aim towards dealing with various handwritings and solving equations with aiming towards the maximum possible accuracy that could be achieved using the various techniques and to find out the most appropriate out of all the proposed techniques.In this study, we initially take a binary image convert it into binary format using preprocessing and eliminating the noise. We use different segmentation and classification techniques have been used to find out the most accurate technique that will give the maximum possible accuracy. We found out that the highest accuracy came in K-Means segmentation and KNN classification technique which are 92.714 and 92.857% respectively.The proposed methodology uses all the techniques of OpenCV.

Automated Machine Learning (AutoML) has progressively established its role in alleviating the complexities associated with traditional model selection and hyperparameter tuning. This research paper introduces a novel amalgamation of AutoML with the benefits of Transfer Learning for image classification [23] through Convolutional Neural Networks [23] (CNNs). By leveraging pre-trained models as a foundation, our framework reduces training time and improves model robustness. Furthermore, a sophisticated early stopping mechanism is integrated, ensuring optimal convergence while mitigating overfitting. The empirical evidence suggests that the fusion of AutoML, Transfer Learning, and Early Stopping paves the way for a new era in efficient and effective image classification, offering a blend of agility and precision.

In Data Science usual exercise is to reiteration throughout several models to observe a best working model. Creating portion of datasets to train and validate model for machine learning to improve performance the model. The splitting ratio of dataset is either 70:30 or 80:20. The problem with this technique is that only one large part is used to train and a small part is used to test ML model. Due to this approach sometimes, model get underfit or overfit. Objective of everyone is always find out the best fil model. CV is a technique which keep a portion of data from the entire dataset and used it for model testing (Validation set), and rest of data other than the part stored to train the ML model. In this paper we apply K fold cross-validation technique with Decision Tree Classifier. We have applied K fold CV by applying distinct K values with Decision Tree Classifier and checking accuracy, precision, recall and F1 value. From different research paper we found that it is difficult to decide the value of K. Our objective is to analyse and identified which value of K is most appropriate. By experimental analysis we found that the accuracy has been improved as compared to the traditional approach. By the observation we found that better for K is 10. BY the average accuracy, precision, recall and F1 value K fold gives better performance for K = 10. Real Life data set has been taken for experimental analysis.

This study explores the potential and application of the newly proposed Forward-Forward algorithm (FFA). The primary aim of this study is to analyze the results achieved from the proposed algorithm and compare it with the existing algorithms. What we are trying to achieve here is to know the extent to which FFA can be effectively deployed in any neural network and to investigate its efficacy in producing results that can be compared to those generated by the conventional Backpropagation method. For diving into a deeper understanding of this new algorithm’s benefits and limitations in the context of neural network training, this study is conducted. In the process of experimentation, the four datasets used are the MNIST dataset, COVID-19 X-ray, Brain MRI and the Cat vs. Dog dataset. Our findings suggest that FFA has potential in certain tasks in CV. However, it is yet far from replacing the backpropagation for common tasks. The paper describes the experimental setup and process carried out to understand the efficacy of the FFA and provides the obtained results and comparative analysis.

Various descriptors are preferred to extract the Local features of the image, including Local Binary Pattern, Local Directional Pattern, and Local Optimal Oriented Pattern. This paper provides the comparative analysis of LBP and Local Optimal Oriented Pattern (LOOP) descriptors for local feature extraction, further used for various applications. While tracking an object from a video, the provided input video sampled into the subsequent frames. For the removal of noise and enhance the frame’s contrast, Median filter is applied on each of the frames. Local features of the image extracted using the Local Optimal Oriented Pattern (LOOP) from these filtered images. The results of LOOP descriptor compared with the Local Binary Pattern (LBP) in terms of histogram and execution time. Experimental analysis shows comparison with the specified feature extraction method in terms of the execution time and accuracy.

Mental health is the main factor which is affected by stress, disease and sarcastic statements or people comments. It effects on persons health directly or indirectly. People cannot share or discuss about their mental condition, even they can’t talk about it. Firstly, they cannot accept that they are suffering mental illness. It is very necessary to predict the mental health of a Pearson in early stage. There is the need to use new strategies for diagnosis and daily monitoring of the mental health conditions. The goal of our research is to develop a module based on feature fusion, which will be performed based on Soergel metric and Deep Kronecker Network (DKN) and early prediction of mental health utilizing Squeeze-MobileNet. It improves accuracy without sacrificing the model efficiency. Particle swarm cuckoo search (PS-CS) is effective and capable to capture the unpredictability of data. We got F1 score and validation score of NN is good as compare to ML.

The classification of Electroencephalogram (EEG) signals into distinct frequency bands is a critical task in understanding brain function and diagnosing neurological disorders. The information obtained from frequency-specific classification has multiple applications, such as frequency-based wheelchair control, frequency-based 36-stroke brain operated keyboard for paralysed patients etc. In this work, a method based on machine learning to develop the frequency-based classification of EEG signals is proposed. The performance of Classical Machine Learning (CML) algorithms and Quantum Machine Learning (QML) techniques for the classification of EEG signals across four frequency bands are investigated. The primary objective is to evaluate the performance of QML models against traditional CML models in terms of computational efficiency, time efficiency and accuracy and uncover potential benefits offered by quantum computing for a particular task of classifying EEG signals. The goal is to assess the advantages of using quantum algorithms for classifying EEG signals. This includes improving accuracy and enhancing efficiency. These findings add to the existing knowledge about how quantum machine learning can benefit neuroscience in terms of enhancing methods that rely on EEG data.

Brain tumor (BT) is a dangerous disease and the process of detecting BT is difficult. Early detection of this disease plays a critical role in protecting the life of humans. Hence, this paper introduced an Adaptive Coronavirus Mask Protection Algorithm (ACMPA)-enabled deep learning technique for detecting and categorizing BT. First, the Magnetic Resonance Image (MRI) brain images are pre-processed using Kalman filtering. After that, BT is segmented by utilizing LadderNet, and the features are extracted which include mean, tumor size, entropy, kurtosis, variance, Haralick texture features, namely Angular second moment (ASM), contrast and Spider Local Image Feature (SLIF). Following this, BT is detected by the Deep Kronecker Network (DKN), where BT is categorized into normal or abnormal. If the detection is abnormal, then BT is categorized into Meningiomas, Gliomas, and pituitary tumors using DKN, which is tuned by the ACMPA. The ACMPA is obtained by integrating the Adaptive concept and Coronavirus Mask Protection Algorithm (CMPA). Furthermore, the proposed ACMPA_DKN acquired the value of accuracy to 90.4%, and obtained the value of TPR and TNR to 91.6% and 92.5%.

This paper introduces an effective algorithm designed for creating AI systems for the Hex board strategy game. The core algorithm, developed, employs the two-distance method for both board evaluation and for sorting of the moves. For empty board positions, the sum of two-distances from both ends is calculated to indicate the position’s weight and is used for sorting. Additionally, the Pattern Search algorithm enhances efficiency by prioritizing moves in crucial regions. The algorithm demonstrated consistent performance across various board sizes, including 7 × 7, 9 × 9, and 11 × 11. When implemented as an Android game, this algorithm maintained excellent performance in the given board sizes.

In today’s dynamically evolving data landscapes, detecting and adapting to concept drifts in streaming data is imperative. Concept drift occurs when there’s a shift in the statistical characteristics of input features, like their mean or variance, or when the relationship between these features and the target label changes over time. This drift can decrease a model’s accuracy because the model is trained on older data. As the data evolves, the model becomes outdated, which can lead to incorrect predictions and reduced performance. This paper introduces the Incremental Restricted Boltzmann Machine (IRBM), an approach designed to address these challenges. The IRBM adapts the traditional architecture and learning paradigms of Restricted Boltzmann Machines (RBMs) to incrementally process and learn from evolving data streams, ensuring model efficacy and accuracy over time. Through extensive experiments, we demonstrate the IRBM’s ability to swiftly detect concept drifts, adapt its internal representations, and maintain robust performance even when confronted with significant data evolutions. The proposed approach outperforms existing methods with an accuracy of 77.42%, 75.32%, 92.12% and 89.21% for electricity, phishing, weather, and rotating hyperplane respectively. Our findings suggest that the IRBM not only offers an effective approach to understanding and adapting to changing patterns in streaming data but also outperforms the other state-of-the-art techniques.

Class imbalance is a salient problem in both machine learning and data mining realms. Sampling techniques have become a cornerstone in solving this challenge, as they enable the creation of class-balanced datasets that is essential for robust model training. Addressing class imbalance not only enhances the predictive accuracy of machine learning algorithms but also ensures fair and unbiased decision-making across various applications, making it a critical aspect of research and development in various sectors. Through this work, we introduce the concept of a generalized oversampling function, unifying existing synthetic oversampling approaches. We explore diverse design decisions for such a function, presenting six functions categorized as linear and non-linear variants. We provide extensive experiments with these functions to gain an in-depth understanding of their behavior. Through our experiments, we observe that the best-performing function is primarily data-driven. Also, it is perceived that non-linear functions like minimum and maximum often depict higher learning capacity and steady performance in comparison to their linear counterparts mainly due to their ability in modelling non-trivial patterns. While moderate input counts would yield desirable performance in these functions, we can see varying robustness from these functions for distorted input data.

These days, financial institutions strive to streamline their loan processes for cost reduction, improved customer satisfaction, and enhanced overall efficiency. Process Mining (PM) offers a data-driven approach that identifies bottlenecks, delays, and unnecessary steps within the loan process. By leveraging event logs from these financial institution processes, PM facilitates optimization and automation, resulting in faster loan approvals. Analyzing event logs can create a comprehensive process model representing the institution’s workflow. This study aims to create a process model specifically tailored for the loan application domain by utilizing the advantages offered by the discovery and conformance steps of PM. The algorithms associated with the discovery and conformance steps are analyzed using two datasets related to the loan application process to identify the most suitable model for the loan application process Optimization. The analysis demonstrates the significance of discovery and conformance algorithms for different quality matrices while generating an effective process model. The proposed methodology reveals that each discovery algorithm comes with its own set of advantages and disadvantages, characterized by varying values of quality metrics. Consequently, the selection of a discovery algorithm is based on the specific quality criteria needed for the task at hand.