Journal of Computing & Biomedical Informatics

A lightweight Approach to Finger Vein Authentication based on Con-trast-Limited Adaptive Histogram Equalization

Hedi A. Guesmi — 2024-10-31

Finger vein recognition is an emerging field that holds a lot of promise for security-sensitive applications. However, most of the past work in this field is dominated by computationally ex- pensive and complex algorithms. In this paper, a new yet simple technique is presented for the preprocessing of finger vein images for biometric authentication. Adaptive histogram equalization, gamma correction, image sharpening, multi-filtering, and contrast adjust- ment constitute the basic steps of this method [AHistGSFC]. For feature extraction, the Histogram of Oriented Gradients [HOG] is used, and the K-nearest neighbour for the recognition. The main contribution of this research is the introduction of a new prepro- cessing algorithm that is not only efficient but also has a very low computational time and has given superior results to deep learning approaches on the same data set. 6 fold cross-validation was used for evaluation. The middle fingers of both hands were found to be the most discriminative, giving 99.06% accuracy and 0.0086% EER. Using all fingers provided in the database, we get 96.96% accuracy and 0.0304% EER. Results were also computed for 3-fold and 2-fold cross-validations for comparison with previous works with that data division. The proposed method gives better results than the latest state-of-the-art algorithms in this field.

A Comparative Study of Advanced Machine Learning Ensemble Techniques for Classification of Breast Cancer

Tayyaba Yasmeen — 2024-09-30

The most prevalent and the life-threatening diseases include Breast cancer worldwide. An early detection and diagnosis which are accurate is essential for improving in survival rates. In the real-time applications, models of machine learning provide powerful tools for aiding in medical professionals in diagnosing the breast cancer very accurately and efficiently. This research focuses on application of advanced ensemble techniques, such as the Random Forest, the Support Vector Machines (SVM), the K-Nearest Neighbors (KNN), the Stacking, the Boosting, and the Blending, to classify cancer of breast as either benign or malignant. After a comprehensive analysis, the Boosting emerged as the highest-performing model with an accurate precision of 0.8493 and a ROC AUC of 0.9051. These findings are align with the Sustainable Development Goal (SDG) 3, which advocates for health care and well-being, that highlighting the importance of an accessible, data-driven education in healthcare and decision of the support systems. Hence, our work suggests that these models can significantly enhance precision in diagnostic, reducing the burden on systems of healthcare, especially in underserved areas.

In Silico Analysis of nsSNPs in the Rb1 Gene for Predicting Pathogenicity and Disease Associations

Anum Munir — 2024-10-15

Single nucleotide polymorphisms (SNPs) are changes at specific spots in DNA. These changes help identify genes linked to diseases or trace inherited conditions within families. Variations in the Rb1 gene can lead to retinoblastoma, which is cancer in one or both eyes, as well as other cancers like osteosarcoma, melanoma, leukemia, lung, and breast cancer. First, this study used the SNP database from NCBI to gather key data. It also analyzed how Rb1 is connected to other genes using GeneMANIA. Ten different tools were applied to screen for harmful SNPs, including SIFT, PolyPhen-2, I-Mutant 3.0, PROVEAN, SNAP2, PHD-SNP, PMut, and SNPs&GO. To estimate conserved amino acid regions, the Consurf Server was used, and Project HOPE was utilized to study the structural effects of mutant proteins. GeneMANIA showed that the Rb1 gene is strongly linked to 20 other genes, such as CCND1 and RBP2. The data obtained from the NCBI's dbSNP indicated that the total number of SNPs within the Rb1 gene region is 36,358. Of these, 345 were found in the 3' UTR, 65 in the 5' UTR and 34,543 in the intron regions. There were 844 coding SNPs including 199 synonymous, 450 non synonmous which consists of 425 missense, five nonsense, and 20 frameshift mutations. The remaining SNPs were of other types. This study focused on the 425 missense SNPs for research. From these, 17 mutations (D332G, R445Q, E492V, P515T, W516G, V531G, E533K, E539K, M558R, W563G, L657Q, A658T, R661Q, D697H, D697E, P796L, and R798W) were predicted to cause harmful effects on the structure and function of the Rb1 protein.

Efficient Electricity Theft Detection Using Hybrid CNN-XGBoost Model

Sayed O. Madbouly — 2024-11-20

Non technical losses especially in distributed networks play key roles in electricity theft that pose serious challenges to power grids. As central electricity is distributed through the power grid to connect all consumers, any fraudulent usage is capable of interfering with the operations of the grid, produce low-quality supply, and even destroy the overall system. This means, as the data volume increases it becomes arduous to identify such fraudulent activities. Smart grids provide a solution in this aspect since electricity flow is bidirectional providing a channel for detection, correction and application of the corrective measures to the flow of the electrical data. Today’s electricity theft detection techniques incorporate one-dimensional (1-D) electric data leading to maximum possible imprecision. This work proposes a model that integrates CNN and XGB known as CNN-XGB. To supplement 1-D theft detection framework, the proposed model includes both 1-D and 2-D power usage data. A comparison with existing benchmark methods, using experimental sample results, shows that the proposed model delivers accurate results for the task, which was the main objective of designing the model.

Brain Tumor Segmentation and Classification Using ResNet50 and U-Net with TCGA-LGG and TCIA MRI Scans

Muhammad Abdullah Aish — 2024-09-29

Brain tumors have become a major source of death in the world. In the case of brain tumor, the brain cells of that particular part grow without any control. The growth has such a serious impact on the normal and healthy cells around the affected part of the brain. Malignancy and benignity are the two types of tumors. Symptoms of the tumor vary according to the place, size, and nature of the tumor. The variable nature of brain tumors is of such a complicated structure that it presents a big challenge for the academics in the field in terms of detection and early classification. A CNN-based model with enhanced “ResNet50 and U-Net architectures” was proposed in this paper. It was used in performing the required analyses on the publicly available “TCGA-LGG and TCIA datasets”. The data in the utilized datasets of “TCGA-LGG and TCIA included that of 120 patients”. The proposed CNN is used, combined with the fine-tuned ResNet50 model for detecting and classifying tumor versus non-tumor images. The model incorporates the U-Net model to precisely segment the tumor region. Accuracy, Intersection over Union (IOU), Dice Similarity Coefficient (DSC), and Similarity Index (SI) metrics are used for measuring the realization of the model. The quantitative results of fine-tuned “ResNet50 report IOU: 0.91, DSC: 0.95, and SI: 0.95”. The combination of U-Net with ResNet50 yielded the best of all, segmenting and classifying tumor regions effectively.

Artificial Intelligence-Enhanced Risk Stratification and Prediction of Cardiovascular Disease through Machine Learning

Faiza Shabbir — 2024-09-30

Since cardiac illnesses provide substantial health concerns to patients, an accurate diagnosis is crucial. Due to the involvement of multiple factors, including smoking, high blood pressure, high blood sugar, excessive cholesterol, and environmental influences, it is difficult to diagnose cardiac abnormalities based only on symptoms. To tackle this, we use innovative machine learning algorithms to evaluate large volumes of medical data, find hidden patterns, and forecast the course of disease. Risk stratification and illness prediction comprise the two main aspects of our study. We evaluate and forecast cardiac anomalies using state-of-the-art algorithms such as Decision Trees, AdaBoost, and Extra Tree classifiers. The main objective of this research is to decrease errors and increase forecast accuracy by merging two datasets. For the old heart disease dataset, we obtained different accuracies for the three classifiers (Decision Tree, 81.5%), Extra Tree, (89%), and AdaBoost, 85.3%). On the newly established cardiovascular disease dataset, the accuracy of AdaBoost (98%), Decision Tree (96.5%), and Extra Tree (99%) was significantly higher. During GridSearchCV optimization, the accuracy rose, demonstrating the robustness of our models. This study shows how individuals who are at a high risk of cardiac events in the future can be identifying using machine learning.

Evaluation of Big Data Tools: A Comparative Study

Prince Hamza Shafique — 2024-10-25

Due to increasing usage of internet huge volume of data is available online. Main source of this gigantic volume of data are social networking sites like Facebook and tweeter etc. It is difficult to handle this huge volume of data. This growing data affects business badly. This data is called Big Data. There are many tools for Big data analytics in this research our focus is on four Big data tools 1) Hadoop, 2) IBM InfoSphere BigInsights, 3) High Performance Computing Cluster (HPCC) and 4) Apache Spark. In this research I have studied architectures, file systems, shortcomings and solutions of those problems. In future this research could be enhanced by running an algorithm on all these tools and then comparing the results. These tools can also be compared by setting some parameters.

Impact of Macroeconomic Indicators on Stock Market Predictions: A Cross Country Analysis

Muhammad Atif Khan — 2024-10-28

This research explores the impact of macroeconomic variables on stock market forecasting across multiple countries, seeking to enhance the precision of predictive models by incorporating essential economic factors. The study utilizes a dataset spanning various economies with distinct financial structures to examine the roles of indicators such as economic growth, inflation, interest rates, unemployment, and currency exchange rates in shaping stock market dynamics. By applying machine learning algorithms and econometric techniques, the research assesses the relevance of these indicators for market predictions and identifies variations across different national and economic contexts. The cross-country approach provides valuable insights into how macroeconomic conditions influence market predictability, offering a comprehensive view on integrating economic variables into forecasting models. The findings contribute to the field by highlighting specific indicators with strong predictive power, enabling investors and policymakers to make more informed financial decisions and adjust their models based on macroeconomic trends. The study concludes by discussing implications for future research in multi-country stock market forecasting and the development of adaptive models that respond to evolving economic environments.

Dynamic Malware Detection in Wireless Networks using Deep Learning

Muhammad Aitzaz Ahsan — 2024-09-29

In the current era of fast digital growth, the significance of security cannot be emphasized enough. Many academics have focused their efforts on creating malware detection systems that utilize data mining techniques to monitor and detect any security breaches. Nevertheless, despite these technological developments, existing systems continue to face challenges in attaining the necessary degree of precision for exact detection. Modern malware employs various evasive techniques, such as polymorphism and metamorphism, to rapidly change and generate numerous variants, challenging traditional detection methods. While machine learning algorithms (MLAs) have shown promise in malware analysis, they often suffer from slow performance due to extensive feature engineering and representation requirements. Advanced deep learning models can eliminate the need for feature engineering but may still face issues with biased performance due to skewed training data, which limits their real-time applicability. This research addresses these challenges by evaluating both classical MLAs and deep learning architectures for malware detection, classification, and categorization. Using a diverse set of public and private datasets, we performed experimental analyses with various dataset splits to train and test models over different timescales. Our key contribution is the development of a novel image processing technique with optimized parameters for MLAs and deep learning models, aimed at improving the effectiveness of zero-day malware detection.

Machine Learning-based Estimation of Soybean Growth

Samia Mishal — 2024-10-19

The ability to determine the yield of soybean crop and the extent of the area of interest with non-destructive methods and at minimal cost is very important because of the increasing world population and the effects of climate change on crop productivity. Yield mapping is important in the crop field at the initial stage of crop management since it gives the total yield in the field, and the distribution of yield in the field which help in the decision making process of which area to fertilize, to irrigate, or to spray pesticide. Thus, the goal of this study was to improve the yield of the soybean crop and the return on investment while utilizing the yield prediction with image data with the least amount of resources and environmental pollution. The Problem Statement is conventional procedures, which are mostly based on the observation and statistical approaches, are often complex, time consuming and prone to errors. That is why the achievement of accurate and timely decisions for crop management is impossible with such restrictions The development of the presented modern technologies implies the search for new approaches that would improve the assessment of the growth with higher accuracy. The proposed research study makes use of Convolutional Neural Network (CNN) to predict the soybean crop yield from an RGB image dataset. The significant phases include data acquisition phase, data pre-processing phase, model-building phase, model-assessment phase, and model-testing phase. The layers in the CNN architecture include the convolution layer, pooling layer, activation layer and the dense layer as well as the output layer. The training of the model was carried in three steps and before each step, the hyperparameters of the model was adjusted. The model is proved to perform well and reliable for yield estimation. The accuracy of testing of the model, is 92.50% and validation accuracy is 94.59% whereas, the training accuracy is 100%.

Analyzing the Impact of Cybercrime and Its Security in Banking Sectors of Pakistan by Using Data Mining

Afsheen Riaz — 2024-11-10

With the rapid digitization of banking services, present day financial institutions are facing threats from cybercriminals. Traditional methods of fraud detection have established inadequate against cyber threats, prompting the adoption of superior technology inclusive of records mining techniques. In this study, we explored the impact of cybercrime on the banking sector in Pakistan by employing data mining techniques. Our dataset consists of a wide range of variables, from cybercrime types and financial losses to customer trust and regulatory compliance. Utilizing classification models, Random Forest with 94% accuracy, Support Vector Machine (SVM) with 93% accuracy, and Decision Tree with 92% accuracy. This research highlights the need for strong security measures in banks to tackle cyber threats. Policymakers and bank professionals can use our findings to make banking safer. Financial institutions can protect their assets and customer information by using fraud detection techniques, thereby increasing trust and confidence in the system in digital banking.

Automated Indoor Plant Health and Pest Control System Leveraging Frontier Technologies for Enhanced Agriculture

Ashar Ahmed — 2024-09-30

The rapid advancement in artificial intelligence and robotics has revolutionized the field of agriculture, particularly indoor farming. This paper presents an innovative solution for enhancing plant health and pest control in indoor farming environments using AI. The system employs a combination of cameras, robotics, and AI algorithms to monitor, analyse, and manage plant health, disease detection, and pest control. This paper introduces an automated system that leverages artificial intelligence and robotics to address these challenges. The purpose is to enhance plant health, identify diseases, and provide automated pest control, thus improving indoor farming outcomes. The system employs a network of cameras moved by a robotic mechanism to capture plant images. AI algorithms analyse these images to determine plant health and identify specific diseases. Automated watering is triggered when soil moisture falls below 45%. Thermal cameras detect pests, prompting automated spray treatments. A web application records and displays plant health, watering schedules, temperature data, pest control percentages, and offers recommendations for timely interventions. It indicates that the system effectively maintains and enhances plant health, detects diseases, and controls pests, thereby increasing the yield and quality of indoor farming produce. The significance of this system lies in its potential to optimize indoor farming, reduce the need for human intervention, and lower the environmental impact of pest control methods. It offers a cost-effective, efficient, sustainable solution for the emerging field of indoor farming.

Neural Network Based Skin Cancer Classification from Clinical Images: Accuracy and Robustness Analysis

Rana Arbab Haider — 2024-10-25

This study aims to investigate the use of neural networks in classifying skin cancer from clinical images, with a specific emphasis on evaluating accuracy and robustness. Skin cancer, melanoma included, poses a significant worldwide health challenge, and early detection is vital for enhancing patient prognoses. Conventional diagnostic approaches heavily depend on clinical expertise, which can be subjective and inconsistent. The progress in deep learning, especially convolutional neural networks (CNNs), presents a promising alternative by automating skin lesion classification with high accuracy. The research involves developing a neural network model using a varied set of clinical skin images, enabling it to distinguish between benign and malignant lesions. Multiple architectures are tested, and their effectiveness is assessed using standard metrics such as accuracy, precision, recall, and F1-score. Beyond measuring overall accuracy, the study emphasizes robustness by evaluating the model in challenging conditions, including variations in illumination, obstructions, and diverse skin tones. Results indicate that neural networks can achieve superior accuracy in skin cancer classification, often outperforming traditional diagnostic techniques. However, robustness remains a crucial area for enhancement, particularly in real-world applications where image quality and patient diversity can fluctuate significantly. By examining the strengths and weaknesses of neural network-based models, this research underscores the potential of AI in clinical diagnostics while highlighting the necessity for further improvements in model generalization to ensure reliable implementation in healthcare settings.

A Novel Multi-Tiered Security Architecture for IoT: Integrating AI, Blockchain, and Efficient Cryptography

Salma Bibi — 2024-10-28

The rapid expansion of the Internet of Things (IoT) has transformed numerous sectors by enabling smart connectivity and data-centric decision processes. Nevertheless, the swift growth of IoT networks poses significant security and privacy challenges due to their scale, heterogeneity, and the substantial amount of confidential information they transmit. This research proposes a layered approach to enhance the protection of IoT devices and their communications. The study explores several key technologies, including artificial intelligence-powered intrusion detection systems (IDS), authentication frameworks based on blockchain, and efficient cryptographic algorithms. The proposed model integrates machine learning techniques such as k-Nearest Neighbors (KNN) and Multi-Layer Perceptron (MLP) to identify and categorize anomalies in IoT data. Results indicate that both MLP and KNN performed exceptionally well, achieving accuracy rates of approximately 98% with minimal.

Real-Time Traffic Flow Prediction using IoT-Driven Machine Learning

Syeda Sitara Waseem — 2024-09-30

Road accidents result in deaths, infections, and many injuries. The main causes of these accidents are traffic congestion, road blockages, and traffic anomalies. Several factors, such as busy routes, damaged roads, or incidents, can trigger traffic. Traffic and roadblocks primarily cause time wastage, energy consumption, delays in reaching destinations, and accidents. Tracking traffic patterns may be a solution to this problem. However, IoT has been successful in a wide range of applications, such as healthcare and inventory management; the downside of tracking traffic is that it is difficult to manage. Therefore, to address this issue, we will design and develop a traffic flow forecasting system using an IoT framework and machine learning. To train and test this system, we will use the ANFIS model, which will then integrated into an Android application. This framework will identify the traffic patterns with the help of IoT sensors in real time, taking into account specific origins, times, peak hours, and speeds. It will then display these patterns in a graphical user interface, enabling users to understand traffic flow pattern and select the most efficient route to reach their destination on time.

Harnessing Artificial Intelligence for Lung and Colon Cancer Classification via CNN

Zulqarnain Iqbal — 2024-09-30

The aggressiveness, strong propensity to spread, and heterogeneity of cancer seem to be the main causes of its very high fatality rate. Throughout the world, lung and colon cancers are two of the most common malignancies that affect individuals of all ages. Accurate and timely detection of these cancers may improve the best aspects of treatment and increase the survival rate. As a complement to the current cancer detection techniques, an extremely exact and computationally effective model is proposed for the rapid and accurate diagnosis of cancers in the lung and colon area. By employing a cyclic learning rate, the accuracy of the proposed techniques is increased while maintaining their processing efficiency. This is easy to use and effective, which speeds up the model's convergence. Furthermore, many transfer learning models that have already been trained are used and compared with the proposed CNN that has attention layers. The validation, testing, and training of the study make use of the LC25000 dataset. It is observed that the proposed model reduces the impact of inter-class disparities between lung adenocarcinoma and lung cancer of squamous cells by offering higher accuracy. By putting the proposed framework into practice, accuracy was improved to 99.04%.

Deep Learning based Smart Healthcare Monitoring System using Sensory Network

Saria Shafi — 2024-11-04

The smart healthcare monitering systems are becoming popular day by day due to failure of traditional healthcare monitoring systems to provide real-time, accurate, and consistent monitoring of persons’s vital signs to provide medical treatment timely. The Internet of Things (IoT) make it possile to design and develop the smart healthcare system to do real-time patient monitoring. IoT plays vital role in monitring systems. This study predicts the person’s health using IoT sensor with advanced Deep Learning (DL) algorithms by consistent monitoring of person’s vital signs, such as heart rate, blood pressure, temperature, and oxygen level. Monitering of vital signs is very important to asses overall health and detect abnormalities The challenge of managing large-scale complex datasets from IoT devices is addressed by DL technology, such as convolutional neural networks (CNNs), recurrent neural networks (RNNs) and artifical neural network (ANN) which analyze the data to detect abnormalities and provide high accuracy. Experimental results demonstrates the effectiveness of proposed approach for person’s health monitering. This research opens up new opportunities to integrate IoT and DL to improve clinical decision-making and patient care.

High-Efficiency Battery Charging for Electric Vehicles: MATLAB Simulation of a Phase-Shifted Full-Bridge DC-DC Converter

Saif Ullah — 2024-09-30

The fast development of electric vehicles (EVs) necessitates efficient charging structures which could meet the growing power demands for advanced using situations. This paper gives a complete MATLAB simulation of a Phase-Shifted Full-Bridge (PSFB) DC-DC converter, aimed at improving the charging efficiency of EV batteries. The proposed PSFB converter architecture focuses on improving power first-class, reducing harmonic distortions, and addressing the challenges related to excessive-electricity EV battery charging packages. By enforcing a segment-transferring manipulate strategy together with the usage of electricity semiconductors and passive components, the simulation evaluates the converter’s overall performance beneath various operating situations, which includes fluctuations in enter voltage and cargo changes. Emphasis is positioned on optimizing efficiency and electricity density to allow quicker and extra fee-powerful charging for EVs. Furthermore, the simulation reviews the consistent-country overall performance, mainly the voltage and contemporary ripple throughout charging cycles, demonstrating the converter's suitability for extending battery life. The outcomes validate the PSFB converter's high charging performance, low energy losses, and progressed power aspect, highlighting its capacity for integration into EV charging networks and different green transportation systems. This examine gives treasured insights into the development of high-performance EV charging infrastructure, specializing in key elements together with reliability, energy nice, and gadget efficacy.

Enhancing Security of Autonomous Vehicles using Layered Strategy with Defensive Techniques-A Survey

Saadia Bano — 2024-10-22

Autonomous vehicles (AV) are a revolutionary advancement in transportation technology defined as independent of humans for their performance. Autonomous vehicles are user and environment-friendly as they are easy to operate and help in traffic flow optimization with other benefits. Autonomous vehicles work with an array of modern technologies like sensors, light imaging detection and ranging (Lidar), cameras, GPS, and advanced computing systems that help the autonomous vehicle to predict its surroundings to make optimal decisions in real time. AV is highly dependent on communication, the base of AV relies on communication like inter-vehicle communication, infrastructure communication, and vehicle-to-everything communication. The high dependency on communication channels attracts adversaries with the possibility of information theft, GPS spoofing, and deployment of malicious software for different fraudulent activities. In this research, we have explored the potential security threats of AV with layered-based model approach. This paper have surveyed the recent research trends with countermeasure strategies. This discussion will not only provide an overview of security challenges of AV but also present the open challenges for further research.

Optimized Deep Learning Framework for Early Detection of Heart Disease

Qandeel Asghar — 2024-10-27

Heart disease is one of the top causes of death worldwide, highlighting the importance of reliable and early diagnostic technologies. This paper provides an optimal deep learning architecture for early diagnosis of cardiac disease, with the goal of improving diagnostic accuracy and efficiency. Our approach uses convolutional neural networks (CNN) and advanced data preparation techniques to analyze crucial patient metrics such as electrocardiogram (ECG) signals, blood pressure, cholesterol levels, and other clinical markers. The model detects cardiac illness with high accuracy, sensitivity, and specificity after rigorous experimentation and optimization, which included hyperparameter tuning and feature selection. The framework is tested on a large dataset, with the findings confirming its robustness and suitability for real-world applications. The suggested deep learning model surpasses previous methods, making it a scalable and effective solution for early detection. This study contributes to the development of automated systems that help healthcare practitioners make timely, data-driven decisions, with the ultimate goal of lowering heart disease morbidity and mortality.

Enhancing Vehicular Network Security: An In-Depth analysis of Machine Learning Approaches

Ezzah Fatima — 2024-12-26

Modern transportation systems heavily rely on vehicular networks, facilitating crucial applications such as autonomous driving, in-car infotainment, traffic management, speed restriction, and road safety. These networks primarily utilize the Vehicular Ad Hoc Network (VANET) architecture, which connects vehicles via roadside units (RSUs) to the edge network and ultimately to a backbone network through wired or wireless connections. However, the open and dynamic nature of VANETs introduces various security challenges that can compromise vehicular communications, potentially jeopardizing the safety and efficiency of intelligent transportation systems. This study examines the current state of security services, common attacks, and application scenarios specific to VANETs, with a focus on machine learning techniques to strengthen these networks. It evaluates advancements, identifies gaps, and suggests future research directions to enhance the robustness and resilience of VANETs in an increasingly connected and automated transportation environment. This study aims to support ongoing efforts to address security issues in VANETs and enable the full potential of vehicular networks in future transportation systems.

Urdu Language Text Summarization using Machine Learning

Imtiaz Khan — 2024-09-29

Text summarization involves creating a concise version of a text while preserving its essential details and core message. This technique allows for quick understanding of lengthy articles, documents, or books, saving time and enhancing comprehension. There are two primary approaches to summarization: extractive and abstractive. Extractive summarization involves selecting key sentences or phrases from the text, while abstractive summarization generates new sentences that convey the original meaning. In the context of Urdu news articles, summarization is particularly valuable, as it enables readers with limited time or attention to grasp the main points quickly. This study explores various methods for summarizing Urdu news articles, evaluating both extractive and abstractive approaches using various datasets. To evaluate the proposed model, we use ROUGE metric which shows the significant improvement and efficiency compared with existing models. The study also highlights challenges and future directions in this field, including the complexity of Urdu sentence structures, addressing biases in content of news, and incorporating the latest development natural language processing and deep learning.

Enhancing Reliability and Sustainability of Green Communication in Next-Generation Wireless Systems through Energy Harvesting

Salahuddin — 2024-09-30

This research investigates strategies aimed at achieving reliable and environmentally sustainable Transmission in future-oriented wireless frameworks via the integration of energy harvesting technologies. As that demand for wireless connectivity continues to escalate, there is a pressing need to minimize environmental impact. Energy harvesting offers a promising solution by harnessing ambient Natural energy forms like solar, wind, and kinetic to energize wireless infrastructure and devices. The central goal of this investigation is to design robust communication protocols and network architectures specifically designed to operate efficiently on harvested energy. This encompasses the formulation of effective power management techniques, adaptive transmission strategies, and dependable error control mechanisms. These innovations are crucial for maintaining seamless connectivity, even amidst fluctuating energy availability. Methodologically, the research employs simulation models to Measure the performance of wireless networks incorporating energy-harvesting techniques across diverse practical scenarios. Key performance Measures like energy efficiency, reliability, latency, and throughput will be meticulously analyzed to assess the effectiveness of the proposed methodologies. The outcomes of this study are anticipated to significantly advance the field of sustainable wireless communication technologies. By leveraging energy harvesting technologies effectively, the research aims to contribute to the development of greener and more resilient next-generation wireless systems. These systems are poised to meet the burgeoning demands for connectivity while concurrently reducing the overall environmental footprint associated with traditional wireless communication infrastructures.

Dose-Response Association between Musculoskeletal Disorders and Physical Factors among Construction Workers

Maryam Naveed Sheikh — 2024-11-04

Background: Musculoskeletal disorders (MSDs) are a variety of inflammatory and degenerative diseases that affect muscles, tendons, ligaments, joints, peripheral nerves and related blood vessels. These contain syndromes like tendon inflammation and associated conditions (tenosynovitis, epicondylitis, and bursitis) that is most commonly seen in the construction workers with prolonged working hours. Objective: The purpose of the study was to find the dose-response association between MSDs & physical factors among construction workers and to determine the association between their pain severity and working duration. Methodology: A cross-sectional study of 317 construction workers. Workers were randomly selected from different areas of Multan. Nordic questionnaire and NPRS were the data collection tools along with the questionnaire derived from previous studies. Results: No significant association was found between the physical factors and MSDs. 49.6% workers experienced mild pain with working duration of 8 hrs. 8.3% workers whose work duration was 10 hours, experienced severe pain. Conclusion: Physical burden factors don’t have any strong predictor role in MSDs of different regions of body among construction workers. Working duration has significant association with pain severity.

Ocular Diseases Detection Using Machine Learning, Deep Learning and Artificial Intelligence Based Techniques

Sayyid Kamran Hussain — 2024-09-30

Glaucoma is one the most common and rapidly increasing eye disease. Glaucoma is a condition which affects the retina and is the most common reason for blindness. Glaucoma cannot be detected in its initial stages as it does not show any of its symptoms. Glaucoma was estimated to affect 60 million people in 2010. In 2020,the glaucoma disease affects around seventy-six million people more, which is expected to rise to 111.8 million by 2040. Early diagnosis and treatment is necessary. Along with expert doctors and health professionals, computer aided techniques will be more useful for early and accurate diagnosis and certainly a great help for the medical professionals.Hence, there are many techniques such as deep learning, machine learning and artificial intelligence techniques to detect glaucoma. For glaucoma classification and identificationthereare different deep learning modelsthat have been reviewed in this work which are Inception-V3, Vgg-16, ECNET, Convolution Neural Network (CNN), Deep-Belief Network, EffcientNet and UNet++ models. Machine learning models have also been reviewed in this work for glaucoma diagnosis which are LSSVM(Least Square-Support Vector Machine), XGboost model, Fundus and OCT, SVM.To the best of our knowledge, this is the only comprehensive study which encapsulates various computer-vision based techniques for glaucoma disease detection.

Optimized AI-Driven Intrusion Detection in WSNs: A Semi-Supervised Learning Paradigm

Syed Shahid Abbas — 2024-10-19

In this study, we have developed an advanced semi-supervised learning model specifically designed to identify four distinct types of attacks in Wireless Sensor Networks (WSNs): Denial of Service (DoS), Probe, Remote to Local (R2L), and User to Root (U2R). Our model leverages the combined advantages of supervised and unsupervised learning approaches, employing a Support Vector Machine (SVM) for the supervised aspect and Density-Based Spatial Clustering of Applications with Noise (DBSCAN) for the unsupervised component. We rigorously tested and validated our model using the NSL-KDD dataset, which highlighted its strong performance metrics, including accuracy and F1-score. Additionally, our research investigated the sensitivity of DBSCAN parameters and their effects on model accuracy, underscoring the importance of precise parameter tuning to achieve optimal results. A notable advantage of our semi-supervised approach is its capacity to manage large amounts of unlabeled data effectively, a challenge that purely supervised or unsupervised methods often face independently. By efficiently utilizing labeled data and integrating clustering techniques, our model shows improved accuracy and effectiveness in detecting intrusions within WSNs. Overall, this research advances the field of intrusion detection in WSNs by introducing a practical and effective semi-supervised learning framework. This framework enhances detection performance across various attack types and provides valuable insights into optimizing model performance through parameter sensitivity analysis and strategic dataset use.

A Trust-Driven Optimization of Role-Based Access Control in E-Health Cloud Systems

Hedi A. Guesmi — 2024-11-21

In today’s world all services related to every sector have made a massive change, especially in health care IT has brought about revolutionary advancement in e-Health. E-Health can only be optimally effective with the integration of Implementing it within a cloud-based infrastructure environment. However, while applying this approach has several benefits, it also raises questions concerning privacy and security. There are inevitable situations when process performance has deteriorated along with the growth of the user base of the Electronic Healthcare Systems (EHS). In response to this challenge, this research presents a novel trust-based access control (AC) mechanism known as Role-Based Access Control (RBAC). It assesses the activity of the user and assigns the position according to the result. The AC module is implemented with SQL server as the back-end administrators may influence user roles and access to EHS several modules. To confirm the trust levels exhibited by users, a.NET-based framework is utilised. The developed new e-Health management framework also confirms the highly secure protection of user data and statistics which are protected against intrusions and other threats.

Predictive Machine Learning Models for Early Diabetes Diagnosis: Enhancing Accuracy and Privacy with Federated Learning

Zill E Huma — 2024-09-29

Millions of people in the world are affected by diabetes, which is a serious chronic illness that have need of early detection for effective management and treatment. Even if they work well, typical techniques for detection are normally very expensive, time-consuming, and invasive. In this regard, machine learning (ML) has become a ground-breaking method for diabetes detection, providing an exact, effective, and non-invasive substitute. We are using the 27,690 instances and nine attributes of the Kaggle diabetes dataset, a combined machine learning model is presented in this paper. We utilized three machine learning algorithms: XG Boost (XGB), Naïve Bayes (NB), and K-Nearest Neighbours (KNN). XGB had the finest accuracy, coming in at 90%. To improve model performance while defending data confidentiality, our methodology includes data collection, pre-processing, training, testing, and parameter adjustment inside a united learning framework. The findings validate machine learning's marvellous potential for enhancing diabetes diagnosis, simplifying early intervention, and lowering medical expenses. Federated learning's integration further keeps patient privacy and data safety, giving it a solid option for extensive clinical use. This work opens the door for more accurate, effective, and accessible healthcare resolutions by highlighting the crucial implication and effectiveness of ML based diabetes prediction.

Detection of DDoS/DoS Attack Methodologies in Cloud Computing Network: A Survey

Tariq Mehboob — 2024-09-30

The concept of cloud computing has been proposed for many years, and the first cloud computing service was launched in the 21st century. Cloud computing uses the Internet or the cloud to provide computing services such as servers, storage, databases, networks, software, analytics, and intelligence. The three types of cloud computing include IAAS, PAAS, and SAAS. CIA, which stands for Trust, Integrity, and Availability, is the foundation of information security. While DDoS/DoS destroys availability, network unavailability and network inadequacy affect reliability. DoS and distributed DoS (DDoS) attacks have become more sophisticated and cannot be stopped by traditional protection tools in cloud computing. Machine learning and neural networks are subcategories of software computing techniques that can be used for network analysis to detect patterns in the occurrence of DDoS/DoS attacks. With the rise of cloud computing, the threat of attacks has also increased. Attackers are also using new technologies to attack the cloud to disrupt services. These attacks can cause serious damage to cloud service providers and their customers. DDoS/DoS attacks attempt to prevent loss of revenue, reputation, and trust. By using cloud computing technology, cloud service providers can ensure that their services are secure and reliable, thereby enhancing confident and proud customer experience.

Artificial Intelligence of Things (AIoT) for Cardiac Healthcare: A Real-Time Monitoring Solution

Sanam Nayab — 2024-10-31

Cardiovascular diseases are one of the most common causes of death worldwide, and it often occurs very suddenly without any chance for medical intervention. These kind of scenarios can be mitigated by proactive monitoring. The current cardiac monitoring solutions for Cardiac IoT groups are typically limited to a set of parameters to be tracked, which fails in integrating AI and often requires physical presence during health data tracking while most patients prefer being checked from their homes so that they can avoid the hospital environment especially during global pandemics like COVID-19. Currently, in Pakistan there is no AIoT (Artificial Intelligence of Things) integrated system for a real time cardiac monitoring. This study introduce AIoT-enabled Cardiac Healthcare Monitoring System, intended for real-time monitoring remotely of vital cardiac parameters. It uses Heart Rate (HR), Oxygen Saturation (SpO₂), Body Temperature and Blood Pressure as key parameters along with presence of a Electrocardiogram signal to provide intelligent monitoring and predictive capabilities. Furthermore, the system generates alerts for anomalous readings and sends data on a cloud server so doctors can access patient history anywhere in the world – ensuring timely medical intervention and cost-effectively enabling healthcare delivery to vulnerable populations far from urban hospitals.

AI-Powered Phishing Detection and Mitigation for IoT-Based Smart Home Security

Noor Fatima — 2024-10-30

Many people with no technical expertise are implementing smart home technology as it becomes more widely available without fully comprehending the privacy and safety risks involved. We interviewed with 40 smart home users with the goal to learn about their concerns and methods for reducing risks in order to fill this knowledge gap and offer knowledgeable advice. According to our research, users have a variety of concerns which are frequently measured against the benefits they observe in smart home technologies. Although some users expressed their concerns, others demonstrated that they were willing to take some risks. But we observed that these concerns weren't usually followed by robust mitigating strategies, mostly because users had restricted technological knowledge or few possibilities. Our study's distinctive emphasis on user experience reveals significant differences between knowledge and application of security and privacy protocols. This research offers useful recommendations for improving user regulate and create IoT-based technologies to Improved security for intelligent houses. Upcoming efforts to enhance the privacy and security of smart home devices can profit from these insights.

Calculating Topological Indices of Benes and Butterfly Network

Kashif Aslam — 2024-09-30

A graph is numerically represented by topological indices. These indices are crucial for topological indices because they affect the quantity structure property connection and the quantitative structure-activity relationship. In parallel computing, digital signal processing, communications, data centers, and network-on-chip design, benes are utilized. In this article, we calculated the quadratic-Contraharmonic index (QCI), contra harmonic-quadratic index (CQI), geometric quadratic index (GQI), quadratic geometric index (QGI), arithmetic Contraharmonic index (ACI) and Contraharmonic arithmetic index (CAI) for the cylindrical Benes network and for both horizontal and vertical and Butterfly network. We use MATLAB tool to give study these networks graphically and draw the comparison bar graphs.

Unveiling Complex Scenes: A Deep Belief Network and Semantic Segmentation Approach

Adnan Ahmed Rafique — 2024-09-29

Scene classification is a meaningful and challenging research field in computer vision due to the wide variety of objects present in a scene, their internal relationships, and inter-class similarities. Thus, complex scene recognition and understanding are needed in various applications, including virtual reality-based scene integration, robotics, autonomous driving, and tourist guide systems. Therefore, a novel scene recognition system that integrates various components to recognize the scenes in complex imagery is developed. Compared to the state-of-art systems, our system combines many significant features for improving the classification accuracy. Initially, the images are acquired and preprocessed. It is worth mentioning that semantic segmentation approaches are powerful as they not only detect objects present in an image but recognize the boundaries of each object. To leverage the effectiveness of semantic segmentation, we propose a modified fuzzy C-means (MFCM) segmentation method that partitions the image into various objects to label the pixels according to different segmented objects. Then, convolutional neural network (CNN) features, the dynamic geometrical (GF), and, blob features (BF) are extracted and fused for further analysis to recognize the scene through a deep belief network (DBN). The latter incorporates a genetic algorithm that optimizes the number of hidden units based on the error rate and the training time. The effectiveness of the proposed system is validated over Pattern Analysis, Statistical Modeling, and Computational Learning Visual Object Classes (PASCAL VOC 2012) and the Microsoft Research Cambridge (MSRC) datasets by achieving 93.30% and 92.53% recognition accuracies respectively.

Machine Learning-Based Multi-Factorial Genetic Disorder Prediction System

Maya Bint Yousaf — 2024-11-11

A genetic disorder is a medical illness caused by an error in a person's DNA. Genes are hereditary units that carry instructions for the body's development, functioning, and upkeep. Mutations or alterations in these genes can cause genetic illnesses, affecting how the body's cells create proteins or carry out certain functions. This study proposed the multi-factorial genetic disorder prediction system using machine learning algorithms specifically Decision trees and Naïve Bayes. The study covered three diseases cancer, cystic fibrosis and diabetes. Diabetes, cystic fibrosis, and cancer are frequently the result of a complicated interaction of genetic, environmental, and behavioural variables. This proposed model can identify those who are more likely to develop certain diseases, allowing for early intervention and personalized preventive care. These proactive healthcare methods not only improve patient outcomes but also contribute to the general efficiency and efficacy of the healthcare system, supporting a change towards more targeted and efficient healthcare delivery.

Detection and Analysis of Brain Tumors on the Basis of their Area and Density by Segmentation

Neelam Shahzadi — 2024-10-30

Brain Cancer is recognized to be a deadly and most prevalent disease around the globe. The prime step in curing brain tumor is its detection, as it is required for diagnosis of this disease. With the help of Computer-Aided Diagnosis (CAD), the detection and diagnosis of brain tumors can be automated. The major issues that are encountered in designing these automated diagnosis systems are efficiency and accuracy. The tumors in Brain Magnetic Resonance imaging (MRI) may be visible clearly; however, the quantification of the tumor affected sites is required. In this regard, computerized image processing methods can provide great assistance. In this paper, the brain tumors have been identified and classified in two major types i.e., malignant and benign tumors, depending upon the texture and shape of the MRI image tumor. Four steps have been followed including preprocessing, skull stripping, segmentation and feature extraction. MATLAB image processing toolbox has been utilized to implement the approach. The results can conclude that shape features and texture of brain tumor in MRI images can be used for their classification with great degree of accuracy.

Diazotrophs-Assisted Phytoremediation of Pesticides: A Novel Approach

Kesho Lal — 2024-09-30

Pesticides have hazardous effects on environment, human health, crop productivity, plant growth and metabolism and they can remain in the atmosphere for longer periods of time due to their persistent nature. Remediation of pesticides is absolutely crucial. Phytoremediation is a newly developed technique utilized in several approaches for the purpose of remediation. Research is being conducted to improve this plant-based technology's efficacy. In this way, symbiotic and rhizospheric microbes are essential to the bioremediation of pesticides. A summary of the data from the ongoing research was used to determine that helping diazotrophs can improve the effectiveness of phytoremediation. Diazotrophic bacteria are essential for pathogen resistance, phytohormone production, nutrient uptake efficiency, biological nitrogen fixation, and degradation of pesticides. They also reduce the need for chemical inputs in sustainable agriculture by enhancing plant health. Diazotrophic bacteria are well known for their biological nitrogen fixation (BNF), biosynthesis of auxins like Indole-3-Acetic acid (IAA), ability to release phosphate from biologically inert forms in the environment for plant uptake. Degradation of pesticides by microbes happens in three phases. First stage reduces the toxicity of pesticides and involves oxidation-reduction mechanisms. Second phase includes conjugation of amino acids and in the last stage secondary conjugation occurs that completely metabolizes the substances. This review's objective is to demonstrate how diazotrophs may support pesticide phytoremediation in polluted soils. An effective combination of diazotroph and phytoremediation technology is suggested by the innovative current review of literature.

A Comparative Study of Machine Learning Methods for Optimizing Mushroom Classification

Muhammad Arslan — 2024-09-30

The exact characterization of mushrooms as consumable or noxious is a basic undertaking that has suggestions for general wellbeing and security. Utilizing machine learning to further develop the forecast precision in such double grouping undertakings addresses a critical progression in both food handling and botany studies. This paper investigate the use of ensemble learning strategies to improve prescient execution utilizing a cleaned and pre-handled form of the notable UCI Mushroom Dataset. The dataset incorporates highlights, for example, cap diameter, gill color, and stem shape, each adding to the characterization precision of mushrooms. In this review, we applied a few noticeable gathering techniques: Random Forest, Gradient Boosting, AdaBoost, Extra Trees, and Bagging. Every procedure uses an alternate methodology to total expectations from different models to work on the dependability and exactness of forecasts over utilizing a solitary model methodology. The assessment of these procedures was directed utilizing a complete arrangement of measurements including accuracy, precision, recall, F1 score, ROC AUC, Matthews Correlation Coefficient (MCC), and Cohen’s Kappa. The Additional Trees group strategy showed prevalent execution, accomplishing the most noteworthy accuracy of 99.17%, precision of 99.20%, and recall of 99.28%. These outcomes were joined by a F1 score of 99.24% and a ROC AUC of 99.94%, demonstrating exceptionally solid prescient capacities. . Such discoveries feature the capability of gathering techniques in basic applications where the expense of misclassification can be serious. The review not just reaffirms the worth of gathering learning in complex order errands yet in addition gives a guide to additional examination into its applications in different spaces requiring high-stakes navigation.

Machine Learning-Based Classification Algorithms for Predicting Hepatitis C: A Comprehensive Analysis

Amna Zulfiqar — 2024-10-28

A disease's accurate diagnosis is one of the most important tasks in the medical field. The most dangerous sickness, which continuously affects a many individuals is hepatitis disease, hence there is a need to automate the disease diagnosis. This study evaluates the diagnostic performance in terms of various parameters and optimization techniques using a range of machine-learning algorithms on a hepatitis dataset. A large dataset that included clinical history, lab test results, and demographic data was used. To get the data ready for analysis, preprocessing techniques such as data cleaning, data discretization, and data normalization were used. The algorithms included in this study are Multi-Layer-Perceptron, Support Vector Machine, Naive Bayes and Random Forest, these algorithms were trained and assessed using metrics including accuracy, recall, precision and F1 score. To minimize overfitting, the model's performance was checked using K-fold cross-validation. ReLU activation function was applied to Multi-Layer-Perceptron for solving the vanishing gradient problem. The classification accuracy scores demonstrate promising outcomes, with SVM scoring 91.86%, NB scoring 89.43%, RF scoring 89.43% and MLP scoring 92.68%. Among all algorithms MLP shows highest frequency.

Diabetes Prediction Using Deep Learning: A Comprehensive Approach Utilizing Feature Selection and Deep Neural Networks

Adnan Shafqat — 2024-09-29

Diabetes is a disorder that has a significant impact on world health. In order to properly treat the illness and avoid complications, early identification is crucial. This paper presents a novel scheme for diabetes prediction based on Ant Lion Optimization(ALO)-enhanced deep learning feature selection. We conducted thorough data processing, able to handle missing values, specifying outliers, and validating the Pima Indian's diabetes-relevant data. The selection of pertinent features was optimizedthat use ALO, and the resulting deep neural network (DNN) was then providedwith classification training. The suggested model outperforms typical machine learning (ML) approaches, with an astonishing 96.50% accuracy. This prediction precision demonstrates the aim to expand predictive accuracy by integrating metaheuristic systems with DNNs. According to our findings, this technique is ideal for dramatically enhancing early diabetes diagnosis and delivering valuable knowledge for medical decisions

Prevalence of the Trigger Finger in Barbers of Multan

Ansa Zoha — 2024-09-30

Background: Trigger finger, medically termed stenosing tenosynovitis. It is characterized by the bothersome occurrence of catching, popping, and locking sensations in the affected finger or fingers. Objective: This study aims to assess the prevalence of trigger finger within the barber community. Methods: The research employed a cross-sectional design. Results: The findings revealed that the incidence of trigger finger among barbers stood at a mere 1.7%. Of the 286 participants. Conclusion: In conclusion, this investigation determined that the occurrence of trigger finger among barbers is notably low at 1.7%. Among the participants, two cases of trigger finger were definitively diagnosed, and three cases showed uncertain indications. The study emphasized the connection between trigger finger and the repetitive hand movements and the prolonged use of heavy equipment inherent to the barber profession.

Deep Learning-Based Methods for Brain Tumor Segmentation: A State-of-the-Art Review

Meiraj Aslam — 2024-11-10

Hospitals have lately begun using machine learning to expedite the diagnostic and analysis process. Now that they have assistance with diagnosis, doctors can expedite the start of the healing process. AI in healthcare may be used for simple to complex tasks in the future, such as phone answering, reviewing medical records, trending and analytics in primary care, computer design and therapeutic medicine, reading radiology images, creating treatment and diagnosis plans, and even having conversations with patients. Medical imaging such as CT, MRI, and X-ray pictures may be interpreted using deep learning models to establish a diagnosis. Inconsistencies and dangers can be identified by the algorithms in the medical imaging data. Cancer detection frequently makes use of deep learning. Brain tumors must be correctly segmented using MRI images in order to aid in clinical diagnosis and therapy planning. However, the lack of certain diagnostic procedures in MRI images makes medical practice more challenging. The recommended method performs better when comparing the quantitative and qualitative results of medical image analysis as it is currently performed. When it comes to the accurate identification of malignant lung nodules in the event of lung cancer detection, CT scans of the chest perform better. Early detection of lung cancer is crucial for patients' chances of survival. Using sparse chest computed tomography (CT) data from earlier research, create a multi-view knowledge-based collaborative (MV-KBC) deep model to distinguish between benign and malignant nodules. However, the MV-KBC model had more accuracy. Nevertheless, the model can only be used to supervise image data. In this research, we present a novel deep learning-based multi view model to alleviate the model's shortcoming. The accuracy of the suggested model was significantly improved, and computation and classification times were decreased, for semi-supervised medical image applications.

YOLOv9-Based YOLO-Enhanced Smart Glasses for Real-Time Recognition of Pakistani Currency: Empowering the Visually Impaired

Mamoona Khalid — 2024-09-30

People who are blind or visually impaired may find it difficult to manage daily tasks, particularly when it comes to knowing the different denominations of cash. We provide an innovative approach to tackle this problem, enabling real-time cash detection through the combination of wearable technology and deep learning. In this work, a deep learning model-integrated smart glasses system powered by a Raspberry Pi IoT device is introduced. The cutting-edge YOLOv9 algorithm is used by the system to accurately recognize cash notes. Training the model included using an extensive dataset of 3,611 photos with seven distinct rupees denominations: 10, 20, 50, 100, 500, 1000, and 5000. The smart glasses immediately alert the user of the denomination when they detect a bank note by means of voice feedback. Our technology boosts the transaction experience and gives visually impaired people more independence in their financial operations, all while maintaining an accuracy rate of up to 95%. As a practical and dependable instrument for daily transactions, the system's lightweight and portable design guarantees simplicity of use in a variety of environments. Providing a reliable and effective method for cash identification, this research marks a substantial breakthrough in assistive technology.

Sentiment Analysis Classification of ChatGPT Tweets Using Machine Learning and Deep Learning Algorithms

Imtiaz Hussain — 2024-09-30

Sentiment analysis is essential for understanding public opinion and emotional responses to specific topics. In this study, we conduct sentiment analysis on a dataset comprising tweets related to ChatGPT. The dataset includes two primary columns: tweets and sentiment labels (positive, negative, and neutral_l). We developed and evaluated machine learning (ML) models to classify these tweets' sentiments. To preprocess the data, we applied standard text cleaning techniques such as removing special characters, tokenization, and stop word removal. The textual data was converted via Count Vectorizer to numerical features, and the labels were encoded using Label Encoder to transform categorical sentiment labels into numerical values. The Convolutional Neural Network (CNN) captured sequential patterns within the tweets and achieved a noteworthy accuracy of 88.25%. The Long Short-Term Memory (LSTM) network has captured temporal dependencies and yielded an accuracy of 89.24%. Logistic Regression (LR) achieved an accuracy of 85.74%, while Decision Tree (DT) and Multinomial Naive Bayes (MNB) models achieved 71.60% and 67% accuracy, respectively. The results demonstrate the efficacy of machine learning models, particularly CNN and LSTM, in accurately classifying the sentiment of ChatGPT-related tweets and effectively capturing sequential and temporal characteristics of social media text, offering insights into public sentiment towards ChatGPT. Our findings have practical implications for understanding user feedback on ChatGPT to enhance its performance and user experience on social platforms.

A Novel Approach to Vitiligo Diagnosis using Artificial Neural Networks and Dermatological Image Analysis

Muhammad Usman — 2024-10-27

This study presents a novel approach to diagnosing vitiligo through the use of artificial neural networks (ANNs) and dermatological image analysis. Leveraging advanced image processing techniques, we analyzed skin lesion images to identify vitiligo with greater precision and speed. Our approach utilizes a pre-trained convolutional neural network (CNN) model, fine-tuned on a dataset of dermatological images to extract critical features from the lesions. The ANN then processes these features to classify the presence or absence of vitiligo. By incorporating patient demographic data along with image analysis, we improved the diagnostic accuracy of the model. This method demonstrates significant potential in reducing diagnostic error and aiding dermatologists in clinical decision-making. The results show improved prediction performance and offer a more efficient, non-invasive alternative for diagnosing vitiligo, with implications for future clinical applications and automated dermatological analysis.

Performance Evaluation of Various Optimizers for Breast Cancer Diagnosis Using Neural Networks

Yasir Nawaz — 2024-09-30

Breast cancer is one of the leading causes of premature death for women worldwide; therefore, early and accurate detection is critical to improving patient outcomes.This study analyzes the effectiveness of multiple neural network optimization algorithms in the classification of breast cancer using clinical data. To assess optimization tools like Adam, Nadam, Adagrad and RMSprop with neural network ,we used a number of efficiency measures,such as accuracy, precision, recall, F1-score, specficity, and ROC-AUC. Our trials, executed at various folds, highlight the positive aspects and drawbacks of each optimizer in relation ti the diagnosis of breast cancer. The results show that Adam Consistently achieves higher balanced accuracy and accuracy than other optimizers. Adam specifically attained a balanced accuracy of 94.12% together with a high accuracy of 94.9%. This research mapped using SDG-3. Our research sheds light on the most effective optimization techniques for creating credible breast cancer diagnosis models.

Assessing the Effectiveness of Ensemble Learning Models for Hepatitis C Detection through Advanced Machine Learning Techniques

Fatima Zafar — 2024-09-30

This paper investigates the potential benefits of utilizing advanced methods of machine learning to enhance Hepatitis C diagnosis tools. We used a publically available dataset to test different ensemble learning techniques, such as Grid Search and Random Search to optimize the parameters of Random Forest, Gradient Boosting, Bagging, XGBoost, and stacking. We evaluated the performance of the model using Cohen's Kappa, F1 score, accuracy, precision, and recall. With 92.37% accuracy, 83.85% precision, and a 70.17% F1 score, XGBoost with Random Search demonstrated the best performance. The results show that medical diagnostics can be improved and that methods of ensemble learning are useful for early Hepatitis C identification.

Prevalence of De Quervain’s Tenosynovitis and its Association with Overwriting Pain and Handheld Devices Usage among College Students in Multan

Ayesha Rani — 2024-11-04

De Quervain Tenosynovitis is an inflammatory and degenerative disease. Thumb function impairment, radial wrist pain, and thickening of the ligamentous structure covering the tendons in the wrist's first dorsal compartment are the hallmarks of this disorder. Moreover, Repetitive strain injury can lead to inflammation and bump of tendons. Overuse injuries of the hand and wrist can affect any of the constituent bones, tendons, ligaments, nerves and cartilage. The purpose of the study was to determine the frequency of De-Quervain’s tenosynovitis, its relationship with the handheld devices usage among college students in Multan. It was a cross-sectional study of 339 College Students conducted on Thumb Problems. Students were randomly selected from different areas of Multan. Prevalence of De Quervain tenosynovitis among 339 students was analyzed among whom 123 students were Positive after performing Finkelstein test. The association between writing and Finkelstein test shows significant relationship. This study shows 36.2% preponderance of De Quervain tenosynovitis among college students in Multan. Pain upon writing and Finkelstein test show positive association.

A Comprehensive Survey of IoT Threats Assessment and Mitigation Strategies

Salman Mushtaq Qureshi — 2024-09-30

As the quantity of gadgets linked to the internet rises, cybersecurity has emerged to be an essential topic especially with the emergence of the IoT. It is crucial to shield IoT systems from numerous threats to avoid possible vulnerabilities and preserve the users’ security. This survey aims at revealing the sphere of influence of the IoT which is a technique that makes it possible for messages to be sent across different physical items.. Used massively in industrial and social contexts, IoT improves ease of life while posing significant risks to security. As IoT devices feature an autonomous functionality and rather limited human intervention, they require smart and secure design. further more, through paper explains the security risks associated with IoT and gives relative attention about different layers and need for effective solution. Through pointing out the issues of portability, resource limitation, and open environments, this work is intended to help researchers and manufacturers strengthen the IoT devices against possible attacks and contribute to the proper, secure, and productive IoT development.

A Comparative Study of Machine Learning Models for Heart Disease Prediction Using Grid Search and Random Search for Hyperparameter Tuning

Suman Arshad — 2024-09-29

An important global health concern is heart failure, for which early detection can greatly improve patient outcomes. Machine learning has proved to be useful in predicting the likelihood of heart disease by looking at factors like age, high blood pressure, and cholesterol. This study compares popular machine learning models, such as Random Forests, Gradient Boosting, Stacking, KNN, SVM, and Logistic Regression. We utilized a Grid Search as well as Random Search to improve the models' efficiency and perform-ability. Following model tuning, the models were determined using metrics like accuracy, recall, F1 score, and precision, AUC, Cohen's Kappa, and MCC. With grid search accuracy of 94.95% and random search accuracy of 94.54%, Random Forest produced the best results. This highlights how important it is to select the right model and adjust its parameters for the best results, and it also shows how well Random Forest predicts heart disease.

Using Blockchain Technology to Enhance Security in VANET: A Comprehensive Analysis

Rania Naveed — 2024-12-26

Vehicular Ad-hoc Networks (VANETs) are critical for enhancing road safety and traffic efficiency through real-time communication among vehicles and infrastructure. However, VANETs face numerous security challenges, including availability, confidentiality, integrity, and authenticity threats. This paper explores various types of attacks targeting VANETs and discusses how blockchain technology can be leveraged to enhance security. Blockchain offers decentralized, tamper-resistant data storage and consensus mechanisms that can mitigate these threats effectively. This study provides a comparative analysis of VANET attack types, their impact on security goals, and proposes blockchain-based solutions to strengthen VANET security.

Smart Irrigation System for Wheat in Southern Punjab Using IOT-based Wireless Technology

Safira Mustafa — 2024-09-30

Agriculture is an important aspect of Pakistan economy. It employs almost 60% of population and generates one-third of exports per year. State of the art agriculture development highlights modern trends to overcome farmer difficulties faced by implementing traditional methods. The general purpose is to integrate farming on agriculture using IoT technology based on Wireless Sensor Network (WSN). This research work focuses on wheat as cultivation and irrigation. The improvement in wheat production demands special conditions, such as temperature and humidity level 50-60%. The proposed system remotely monitors growth of wheat crops using wireless communication. The monitoring system is based on ESP8266 (microcontroller) and DHT11 sensor to read the temperature and humidity level of crops under remote observation model. Moreover, GSM module transmits message to the farmer regarding environmental parameters impact on wheat crop. If humidity level goes up or down from its optimum range then farmer adopts the precautionary measures to maintain humidity within desired range. This model prevents the crop from environmental losses and increase the productivity of crops in south Punjab region.

Tidal Effect of Modified Schwarzschild Black Hole Under the Influence of Rindler Acceleration and Cosmological Constant

R. Minhas — 2024-09-30

We study the effects of tidal force and the behavior of the geodesic deviation vector in the background of a modified Schwarzschild black hole. In this analysis, we obtain the curvature tensor on a tetrad basis to evaluate the radial and angular tidal force on a radially free-falling particle towards a black hole. The radial tidal forces increase with an increase in the Λ (cosmological constant) value but decrease when the value of r increases. The variation of angular tidal force for Lambda shows increasing behavior and gradually decreases to a smooth curve when increasing the radius. We solve the geodesic deviation equation numerically and analyze how the geodesic separation vector varies with radial coordinates for two neighboring geodesics under suitable initial conditions. All the obtained results are tested for a modified Schwarzschild black hole by constraining the value of the acceleration parameter and the cosmological constant. The results are also compared with a simple Schwarzschild black hole.

Deep Learning Approaches for Brain Tumor Detection and Segmentation in MRI Imaging

Usman Humayun — 2024-10-27

The identification and delineation of brain tumors are essential for precise diagnosis, treatment planning, and improved patient outcomes. MRI has emerged as the preferred imaging method, offering high-resolution scans with detailed brain tissue differentiation. Recent strides in deep learning have significantly enhanced the automation of brain tumor detection and segmentation, diminishing the need for manual analysis. This review examines state-of-the-art deep learning techniques for brain tumor detection and segmentation in MRI, emphasizing architectures such as CNNs, U-Net, and advanced models incorporating GANs. The study explores the integration of these models with various MRI modalities, including T1-weighted, T2-weighted, FLAIR, and contrast-enhanced MRI, to achieve greater precision in tumor boundary and type identification. Furthermore, the paper addresses challenges like data heterogeneity, model interpretability, and computational requirements, alongside recent advancements in data augmentation and model explain ability. This research underscores the potential of deep learning to streamline clinical workflows and support radiologists in early and accurate brain tumor diagnosis, while also considering future directions for enhancing robustness and clinical applicability.