VOLUME 14, ISSUE 8, AUGUST 2025

Abstract: Income prediction from demographic data remains challenging due to inherent class imbalance and the black box nature of modern machine learning algorithms. This study develops a comprehensive explainable AI framework to predict income levels using the Adult Income dataset while addressing the critical 76/24 class distribution skew. The research implements and compares four state of the art algorithms (XGBoost, LightGBM, RandomForest, and CatBoost) enhanced with SMOTE balancing and optimal threshold selection. Through systematic application of SHAP, LIME, and permutation importance methods, the framework provides transparent model interpretability. Results demonstrate that LightGBM achieves the best performance with 72.91% F1 score and 82.23% balanced accuracy after threshold optimization, representing a significant improvement over baseline models. The XAI analysis reveals marital status and capital gains as dominant predictive features, with strong consensus across explainability methods. Learning curve analysis confirms model convergence at approximately 35,000 samples with minimal overfitting gaps below 3%. The framework's novelty lies in combining multiple explainability techniques with systematic threshold optimization for imbalanced data. These findings have important implications for fair and transparent automated decision making in financial services, lending, and human resource applications where understanding model reasoning is crucial.

Keywords: Explainable artificial intelligence, income prediction, class imbalance, threshold optimization, SHAP analysis, machine learning interpretability.

Abstract: Brain tumors necessitate early and precise diagnosis for improved patient prognoses. This project introduces an advanced automated system for brain tumor detection, classification, and staging utilizing MRI imaging and Convolutional Neural Networks (CNNs). The system employs robust image preprocessing and tumor segmentation, followed by deep learning-based classification to identify tumor type (e.g., glioma, meningioma, pituitary), stage (early, intermediate, advanced), and precise spatial location. Rigorous evaluation on public datasets demonstrates high accuracy in detection and classification across key metrics, affirming its diagnostic efficacy. Evaluated on publicly available datasets, the system demonstrated high accuracy in detection and classification, evidenced by strong metrics like precision, recall, F1-score, and overall accuracy. A user-friendly graphical interface (GUI) is also integrated for easy visualization and interpretation by healthcare professionals. Coupled with an intuitive graphical user interface for clinical interpretability, this non-invasive and time-efficient solution significantly reduces diagnostic error and aids in early intervention. This pioneering framework holds substantial promise for revolutionizing clinical diagnostics and treatment planning, with future potential for 3D imaging integration and enhanced model robustness. This automated and reliable solution has significant potential for clinical diagnostics and treatment planning by reducing human error and facilitating early diagnosis.

Keywords: Brain Tumor, MRI, Deep Learning, Convolutional Neural Networks, Image Segmentation, Tumor Classification, Medical Imaging.

Abstract: Phishing attacks pose a serious cyber security threat by imitating legitimate websites to steal sensitive data. This study presents a hybrid phishing detection system integrating Machine Learning (ML), Deep Learning (DL), and Ensemble Learning (EL). Feature selection techniques such as Information Gain, Gain Ratio, and Principle component Analysis (PCA) are applied to extract the most relevant indicators from a dataset of 11,055 URLs. ML classifiers (SVM, DT, KNN), EL models (RF, XGBoost, AdaBoost), and DL architectures (LSTM, GRU, CNN) are used. A hybrid model fuses LSTM and GRU outputs, processed by ensemble classifiers and finalized by a meta-classifier. The model captures both structural and sequential URL features, improving accuracy, reducing false positives, and enabling real-time adaptability. The framework can be deployed in email clients, browsers, or gateways to safeguard users from phishing threats. This scalable and intelligent system outperforms individual models and adapts to evolving phishing tactics, contributing to a more secure online ecosystem.

Keywords: Phishing, Machine Learning, Deep Learning, Ensemble Learning, Hybrid Model, Cyber security

Abstract: This paper presents an intelligent, data‑driven predictive artificial intelligence system for early diagnosis of four major medical conditions: vitamin deficiency, heart disease, stroke, and diabetes. The system integrates image‑based and structured‑data modes of analysis. A convolutional neural network (CNN) processes clinical images—such as ocular, nail, or lingual photographs—to detect signs of vitamin deficiencies. Meanwhile, decision tree and random forest classifiers are trained on structured patient data to estimate the probability of heart disease, stroke, or diabetes. The architecture features a dual‑interface design: a Flask‑based web API handles model inference, data ingestion, and prediction delivery, while a C# Windows Forms application serves as a secure admin console for user authentication, message management (text and multimedia), and integration with the predictive engine. Results indicate the system’s potential for accelerated, non‑invasive screening support.

Keywords: Deep learning, convolutional neural network (CNN), machine learning, decision tree, random forest, Windows Forms application.

Abstract: The emergence of virtual worlds in the Metaverse introduces new possibilities-and demands-for educators designing learner-centered digital experiences. However, a structured framework covering the pedagogical, technological, social and ethical competencies necessary for the educational use of the Metaverse is lacking. This study presents MetaFramEdu, a multidimensional competency framework developed to support educators and educational institutions in the effective and responsible integration of virtual worlds into educational practice. Based on existing digital competency frameworks and recent research in immersive learning, MetaFramEdu defines core and specialized competency areas such as: (1) Virtual Environment Design, (2) Pedagogical Implementation, (3) Technological Dimension, (4) Digital-Social Management, (5) Ethics, Safety and Inclusion, etc. For each domain, performance indicators and proficiency levels are proposed. This paper outlines the theoretical underpinnings, development methodology, framework structure, and implications for teacher education and professional development.

Keywords: Metaverse, digital competence, immersive learning, teacher frameworks, virtual environments.

Abstract: Malaria remains a significant public health challenge in Nigeria, with Bayelsa State experiencing persistent high transmission rates despite control efforts. This study developed a comprehensive deep learning-based malaria forecasting and early warning system for the eight Local Government Areas (LGAs) in Bayelsa State. The system utilizes Long Short-Term Memory (LSTM) neural networks enhanced with Principal Component Analysis (PCA) to predict malaria cases and generate early warnings through 2028. Historical malaria surveillance data from 2019-2024 was integrated with environmental variables including rainfall, temperature, humidity, and vector density indices. The model incorporates sophisticated feature engineering, including lag variables, seasonal indicators, and intervention coverage metrics to capture complex temporal patterns. PCA dimensionality reduction improved computational efficiency by 37% while enhancing predictive accuracy. The LSTM+PCA model achieved exceptional performance with R² = 0.939, RMSE = 14.04, and MAE = 10.02, substantially outperforming traditional approaches including ARIMA (R² = 0.849) and baseline models. Early warning thresholds were established using percentile-based methods, with LGA-specific values ranging from 146.5 to 179.5 cases, enabling localized outbreak detection. Model interpretability was enhanced through SHAP (SHapley Additive exPlanations), permutation importance, and Partial Dependence Plot (PDP) analyses, revealing climate variables and lagged malaria cases as primary transmission drivers. The system provides forecasts extending to 24 months, though accuracy assessment was limited to the test period, demonstrating sustained low-risk classifications across all LGAs through 2028. This innovative approach offers a robust tool for public health authorities to implement targeted, data-driven malaria control strategies, with real-time prediction capabilities under 9 milliseconds enabling integration into existing health information systems for improved epidemic preparedness and response.

Keywords: Malaria forecasting, LSTM neural networks, Early warning system, Bayelsa State, Nigeria.

Abstract: A cycle of hunger and food waste seems impossible to stop. Food Waste, Food Security, and Food Justice are the three principles that govern the Red Cross's work on the relation between food and health [1]. When we consider the three principles, there are several facets to the work we could highlight: from our pioneering research that led to the 2011 publication of "The Road to Zero Hunger: A Global Imperative" (a working paper that lays out the relation of food and health and why we must achieve Food Security and Food Justice) [2]; to our partnerships with universities, government agencies, and NGOs (including work in the United States with Feeding America and the Love Food, Hate Waste initiative) unravelling the complicated causes and consequences of the intersection of food waste and hunger [3]; to the development of Zero Hungry People, the online platform/operating system that is designed for volunteers, supervisors, and the served to achieve the aims of the program. (By the way, "Zero Hungry People" is a title straight from the work of the Red Cross) Key Words: Food Security, Food Justice, Food Waste Reduction, Zero Hunger Initiatives, Community-Based Food Systems.

Abstract: A Cyber hacking breaches and prediction using machine learning is one of the emerging technologies and it has been a quite challenging tasks to recognize breaches detection and prediction using computer algorithms. Making malware detection more responsive, scalable, and efficient than traditional systems that call for human involvement is the main goal of applying machine learning for breaches and prediction. Various types of cyber hacking attacks any of them will harm a person’s information and financial reputation. Data from governmental and non – profit organizations, such as user and company information, may be compromised, posing a risk to their finances and reputation. The information can be collected from websites that can be triggered by cyber-attack. Organizations like the healthcare industry are able to contain sensitive data that needs to kept discreet and safe. Identity theft, fraud, and other loses may be caused by data breaches. The finding indicates that 70% of breaches affect numerous organizations, including the healthcare industry. The analysis displays the likelihood of a data breach. Due to increased usage of computer applications, the security for hosts and network is leading to the risk of data breaches. Machine learning methods can be used to find these assaults.

Keywords: Cybersecurity, Data Breaches, Machine Learning, Malware Detection, Breach Prediction, Cyber Hacking, Anomaly Detection, Network Security, Artificial Intelligence (AI), Intrusion Detection Systems (IDS), Risk Analysis, Identity Theft, Healthcare Data Security, Threat Intelligence.

Abstract: Food fraud costs a lot of money and erodes consumer and merchant confidence. One of the most expensive spices in the world, black pepper is prone to adulteration for commercial gain. Adulteration is frequently caused by economic fraud, negligence, a lack of basic sanitation, or intentional tampering with food. Pepper, once worth its weight in gold, today makes up around 35% of the global spice trade. Because they are readily available, inexpensive, and morphologically like black pepper, papaya seeds are frequently employed as an adulterant in black pepper. Black peppercorns and papaya seeds almost appear identical to the unaided eye, but we can tell them apart using picture processing.

Keywords: Food Adulteration, Black Pepper, Image Processing, Accuracy

Abstract: Humanity has faced a serious threat from COVID-19, the most virulent coronavirus strain. It was deemed a global emergency and pandemic by the World Health Organization (WHO), impacting countries worldwide. Numerous industries have experienced significant disruptions, including business, retail, automotive, finance, and aviation. The most affected by these has been schooling, as kids' regular routines have been disrupted. But in these difficult circumstances, technology has become an indispensable lifesaver. The higher education industry has seen a new normal brought about by COVID-19, which has changed academic procedures, redefined crisis management techniques, and completely changed the online learning environment. E-learning platforms have experienced unparalleled expansion in India, for example, with a 600% increase in training in communication skills and a 280% increase in business fundamentals courses. The benefits and difficulties of e-learning, its platforms, and the lockdown are examined in this chapter.

Keywords: Covid19;E-Learning;Digital Tools; Pandemic; Lockdown.

Abstract: The project focuses on the creation of an application based on the web, entitled "Rural Services Management System" that helps rural communities to reach and administer government schemes and public services easily and administer them. Many people in the villages face problems trying to request schemes or application services, since they do not have clear or easy access to information. This system is designed to solve the problem provided by a simple and easy -to -use platform, where users are able to register, log in, & you can see available schemes and request services without visiting individual government offices. The system has two main users: administrators and users of the city. Administrators can add new plans, the user can see applications and manage services, while users can see plans, request services and track their applications. It also follows a modular structure, which makes the code easier to manage and update. As the project shows how technology will be used to solve real world problems in rural areas by improving communication between the public and the government. It reduces paperwork, saves time and makes public services more accessible to those who need their. This system can also be updated in the future by adding mobile support or language translations to help users even more. In general, it is a useful step towards digital development in the communities of the villages.

Keywords: E-Governance, Rural Development, Public Service Delivery, Digital Inclusion, Scheme Tracking System

Abstract: Prior to the modern days of the digital technology, it is important to manage the finances efficiently. Lots of people are tracking their spending with paper and complex spreadsheets that are confusing and not very easy to enter. To make tracking their budget spending simple, we designed a Expense Tracker Web Application with ASP.NET Core MVC that allows users to enter income and expenses, and categorize spending like (bills, groceries, travel, etc.) and set budget limits for the categories. The results can be viewed in charts summary, and grouping the reports and make it easy to see where spending is occurring, updating to the latest secure authentication methods, we have made the entry as easy and user friendly as possible. The program provides a responsive design, an SQL server embedded database to safely store their information.

Keywords: Data Visualisation, Budget management, ASP.NET Core MVC, Expense Tracker, PDF Export.

Abstract: Working capital management (WCM) is a crucial determinant of a bank’s short-term financial health, efficiency, and long-term sustainability. This study examines the working capital management of HDFC Bank over a decadal period (2015–2025) using data extracted from its published annual reports. The analysis incorporates liquidity indicators such as the Cash-Deposit Ratio (CDR), Credit-Deposit Ratio (CD Ratio), Liquidity Coverage Ratio (LCR), Net Working Capital (NWC), Current Ratio, and Quick Ratio. Findings reveal that HDFC Bank consistently maintained robust liquidity buffers while ensuring efficient credit deployment, balancing stability and profitability. The results further highlight how external shocks, particularly the COVID-19 pandemic, temporarily altered liquidity strategies, leading to higher reserves and moderated lending growth. Overall, HDFC Bank demonstrated resilience and financial prudence, aligning with Basel III liquidity standards and Reserve Bank of India’s regulatory requirements. The study carries practical implications for policymakers, regulators, and financial managers in designing effective liquidity policies that promote stability while safeguarding profitability in the banking sector.

Keywords: Working Capital Management, Liquidity Analysis, HDFC Bank, Financial Performance, Decadal Study (2015–2025), Current Ratio and Quick Ratio, Indian Banking Sector.

Abstract: This project is about building a helpful computer tool to find out if students might have a learning disability (LD) and then give them special advice. First, we collected information about many students, like their age, grades in different subjects (math, reading, English, science), and other things like if they have trouble paying attention or a family history of LD. We used this information to train two computer brains, called machine learning models (Random Forest and Support Vector Machine), to guess if a new student might have an LD. We picked the best brain based on how accurate it was. If our computer brain thinks a student might have an LD, it doesn't just stop there. It then asks the student to take small quizzes in different areas like math, grammar, memory, and how they solve problems. After the student finishes these quizzes, the computer figures out which areas they struggled with the most. For these tough areas, the system then gives personalized suggestions. For example, it might suggest certain yoga poses to help with focus or specific exercises to practice for memory. All the results, predictions, and advice are saved securely. This project is a simple but useful way to help students and their families get a better understanding and find ways to support learning.

Keywords: Learning Disability, Machine Learning, Prediction, Personalized Recommendations, Educational Support, Data Analysis, Random Forest, Support Vector Machine, Student Assessment, Yoga Exercises.

Abstract: Getting sick is something everyone experiences and figuring out what's wrong can sometimes be tough. Imagine having a smart computer system that could help you understand what might be causing your symptoms and even suggest ways to feel better. This project is all about creating such a system, using computer science to help people with their health. We built a system that takes a list of symptoms someone might have, like "itching" or "fever," and then uses powerful computer programs to guess what disease they might have. We used a special kind of data that lists many symptoms and their related diseases. We trained several "machine learning" models, which are like very smart pattern-spotters, to learn from this data. The most successful model, called SVC (Support Vector Classifier), along with others like RandomForest and Gradient Boosting, showed amazing accuracy, correctly identifying diseases almost every time in our tests. After predicting the disease, our system also provides helpful information like what to do to be careful, what medicines might be used, what foods to eat, and even some exercises. This entire system is packaged into a user-friendly website, making it easy for anyone to get quick, preliminary health information and even generate a basic health report.

Keywords: Artificial Intelligence, Machine Learning, Disease Prediction, Healthcare Recommendation System, Symptom Analysis, Medical Diagnosis, Personalized Medicine

Abstract: Initiating and involving oneself in social media has become involved in all people's lives today. It allows people to connect with people in different places and share content, information, experience, ideas, and more. Since so many people take part in these activities every day, businesses and organizations are getting into the market as well by marketing, advertising, promoting their brands, and getting more clients. It is also possible to analyse user activity with post engagement. This work used datasets of different features from the past to understand engagement of post and applied Machine Learning (ML) methods to analyse and interpret user activity and measure the amount of engagement of users. The datasets included caption and post time, media type, post length, CTR (click-through rate), ad-interaction time, and hashtag. The datasets also used Machine Learning (ML) algorithms to predict how many interactions a post may get. The results are displayed in graphs that show cluster of users and the engagement activity of each post.

Keywords: Machine Learning, Social Media, Engagement Metrics, Predictive Analytics.

Abstract: The exponential growth of cyber threats in the digital era has necessitated the development of sophisticated countermeasures that can adapt to evolving attack vectors. This comprehensive literature review examines the application of artificial intelligence (AI) and machine learning (ML) algorithms in developing effective cyber attack countermeasures from 2000 to 2024. Through systematic analysis of 142 peer-reviewed publications, this study identifies key AI/ML techniques including deep neural networks, ensemble methods, reinforcement learning, and hybrid approaches that have demonstrated significant efficacy in threat detection, prevention, and response. The research reveals that while traditional signature-based security systems achieve detection rates of 60-75%, AI-driven solutions consistently demonstrate superior performance with accuracy rates exceeding 95% in controlled environments. However, challenges persist in areas such as adversarial attacks, model interpretability, and real-time deployment constraints. This review synthesizes current methodologies, evaluates their effectiveness across different attack scenarios, and provides insights into future research directions for AI- enhanced cybersecurity frameworks.

Keywords: Artificial Intelligence, Machine Learning, Cybersecurity, Intrusion Detection, Threat Intelligence, Deep Learning, Malware Detection

Abstract: In today's world, traffic jams happen cause there's just too many vehicles on the road, which means ambulances and other emergency vehicles get delayed, and that's a real risk to people's lives. our smart traffic signal system for ambulances using IoT will give priority to ambulances at traffic lights. when ambulances get close, the system spots them with RFID and switches the lights to green, while the rest stay red. Ambulances can pass through traffic lights without jams. In addition, we have developed an android application that ambulance driver can select the patient’s disease type and enter the hospital code which is sent in real time to the hospital dashboard. The app also displays a map with quickest path from the present site to the hospital that will help the driver reach the destination easily.

Keywords: IoT, Ambulance, RFID, Mobile App.

Abstract: The field of healthcare is among the many new application opportunities made possible by the rapid growth of blockchain technology. By doing a comprehensive technical examination, the following paper looks at the benefits and drawbacks of contemporary blockchain technology applied in the healthcare sector. In addition to an overview of open themes, research perspectives and current research problems across all healthcare application fields, a comprehensive study of new blockchain-based healthcare technologies and related applications is also given. This paper provides a complete review of the potential uses of blockchain technology in the medical field, as well as an examination of the ways in which such applications impact healthcare markets and create fresh possibilities for business.

Keywords: Blockchain, Healthcare, EHR, IOT, Blockchain applications, Supply chain management, Health insurance, AI.

Abstract: Diabetes, commonly known as diabetes mellitus, is a condition that affects how the body processes blood sugar. It occurs when the pancreas either cannot produce enough insulin or the body is unable to effectively use the insulin that is produced. Insulin, a hormone secreted by the pancreas, facilitates the transport of glucose from food into cells, where it is used for energy. Uncontrolled diabetes often leads to hyperglycemia (high blood sugar), which, along with other health complications, can significantly damage nerves and blood vessels. According to 2014 statistics, a substantial number of individuals aged 18 and older had diabetes, and in 2019, diabetes alone was responsible for 1.5 million deaths. However, with the rapid advancement of machine learning (ML) and deep learning (DL) classification algorithms, early detection of diabetes has become significantly more feasible across various fields, including healthcare. In this study, we conducted a comparative analysis of multiple ML and DL techniques for early diabetes prediction. We utilized a diabetes dataset from the UCI repository, comprising 17 attributes, including the target class, and evaluated the performance of all proposed algorithms using a range of performance metrics. Our experiments indicated that the XGBoost classifier outperformed all other algorithms, achieving nearly 100% accuracy, while the remaining models demonstrated accuracy levels exceeding 90%.

Keywords: Diabetes prediction; XGBoost; KNN; CNN; LSTM; Classification.

Abstract: Conventional machine learning (ML) algorithms are rapidly advancing with the introduction of novel learning techniques. These models are continuously improving through hybridization and ensemble approaches, enhancing their computational efficiency, functionality, robustness, and accuracy. In recent years, numerous hybrid and ensemble ML models have been proposed. However, a comprehensive survey of these models is still lacking. This paper aims to address this gap by presenting a state-of-the-art review of emerging ML models, highlighting their performance, applications, and categorization through a novel taxonomy.

Keywords: machine learning; deep learning; ensemble models

Abstract: Cancer remains one of the leading causes of mortality worldwide, and its early detection plays a critical role in improving patient survival rates and treatment outcomes. Traditional diagnostic methods, while effective, often face limitations such as high cost, delayed detection, and dependency on expert evaluation. Recent advancements in Artificial Intelligence (AI) and Machine Learning (ML) have opened new possibilities for enhancing cancer diagnosis, prediction, and prognosis. These technologies enable automated data analysis, uncover hidden patterns, and support clinical decision-making with higher accuracy and efficiency. This study explores various AI and ML approaches applied in early cancer detection and prognosis, focusing on supervised learning, deep learning, and ensemble techniques. The integration of algorithms with medical imaging, genomic data, and electronic health records has demonstrated remarkable improvements in identifying cancer at early stages. Deep learning models, particularly convolutional neural networks, have shown promising results in analyzing histopathological and radiological images. Similarly, machine learning algorithms such as Support Vector Machines, Random Forests, and Gradient Boosting have been effective in predicting cancer risk factors and survival rates. The abstract also highlights challenges associated with AI adoption in healthcare, including data privacy, model interpretability, and the need for large, high-quality datasets. Despite these challenges, AI-driven solutions hold immense potential to complement traditional diagnostic practices and advance personalized medicine. Future research should focus on explainable AI, robust validation frameworks, and collaborative systems that bridge the gap between data scientists and medical professionals.

Abstract: Cancer is recognized as one of the most severe health threats, causing millions of deaths worldwide each year. Among its various types, lung cancer stands out as the most aggressive, with the highest mortality rate. Hence, the development of reliable and precise methods for detecting lung cancer is critical to ensure timely and effective treatment. Designing a strong and accurate classification model is particularly important in medical diagnostics. Due to its widespread occurrence and tendency to remain hidden during the initial stages, lung cancer underscores the urgent need for efficient detection and classification techniques. Globally, it is one of the most common and deadliest cancers, making a significant contribution to cancer-related deaths. Its silent progression during early phases often leads to late diagnosis, when treatment options become less effective. Robust classification systems can help bridge this diagnostic gap by identifying subtle and complex patterns in medical images. Positron Emission Tomography (PET) is widely applied for diagnosing and staging multiple cancers, including lung, liver, and lymphoma. Correct subtype identification is vital for tailoring effective treatment strategies. For instance, lung cancer includes subtypes such as adenocarcinoma, squamous cell carcinoma, and small cell carcinoma, while liver cancer can present as hepatocellular carcinoma or cholangiocarcinoma. Likewise, lymphoma has categories such as Hodgkin’s lymphoma and diffuse large B-cell lymphoma. Subtype classification using PET imaging, therefore, carries substantial clinical importance. However, a key limitation in real-world clinical settings is the scarcity and imbalance of subtype-specific datasets. The major challenge, then, is achieving accurate subtype classification when working with limited data.

Keywords: Lung Cancer, Positron Emission Tomography, early-stage manifestation

Abstract: Brain tumours are a significant health concern, and timely and accurate diagnosis is crucial for patient care. Magnetic Resonance Imaging (MRI) is a widely used non-invasive diagnostic tool for brain tumour detection. However, there are challenges in accurately classifying brain tumours from MRI images, including: Image Variability in MRI images can vary in terms of resolution, contrast, and acquisition parameters, making it challenging to develop a consistent classification method. There have been too many methods developed in recent years to diagnose brain tumour. Heterogeneity of Brain tumours come in various types (e.g., glioblastoma, meningioma) and grades (low-grade, high-grade), each requiring different treatment strategies. Accurate classification must account for this heterogeneity. From this study it has been found that identifying and extracting relevant features from MRI images that can discriminate between different tumour types and healthy brain tissue is a complex task. Limited Training Data for the availability of labelled MRI data for brain tumour classification is often limited, and collecting large datasets can be time-consuming and costly. Interpretability, ability to interpret the decisions made by machine learning models in the context of brain tumour classification is crucial for medical professionals to trust and use these tools.Therefore, there is a need to develop a robust and accurate machine learning system that can effectively classify brain tumours from MRI images by addressing the challenges of image variability, heterogeneity, feature selection, limited data, and providing interpretable results.

Keywords: Brain tumour detection, machine learning, MRI, Heterogeneity.

Abstract: Breast cancer is a profound global health challenge for women, where early detection is critical for saving lives, and while mammography is the standard screening tool, the manual interpretation of these images is a complex and often subjective task that can sometimes lead to errors. To address this, machine learning algorithms are now being developed as powerful aids, capable of analysing mammograms with advanced image processing to automatically identify subtle signs of malignancy, and by training these models on extensive datasets, we can create systems that achieve remarkable accuracy, offering a reliable, complementary tool that enhances traditional diagnostics and holds the transformative potential to improve early detection rates and patient outcomes worldwide.

Keywords: Machine Learning, Image Processing, Segmentation, Early Detection, Artificial Intelligence.

Abstract: The rapid proliferation of fake news on digital platforms poses a significant challenge to public trust, social stability, and informed decision-making. To address this concern, this study investigates the effectiveness of conventional machine learning classifiers for fake news detection using hand-crafted textual features. Several widely used models, including K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Logistic Regression, and Random Forest, were evaluated after applying rigorous preprocessing and feature extraction techniques. Experimental results highlight that KNN and SVM demonstrated superior performance, achieving up to 88% accuracy in distinguishing between authentic and fabricated news. The findings underscore the importance of leveraging well-structured datasets and robust classification techniques to combat misinformation effectively. This work provides a foundation for developing scalable and reliable automated systems for mitigating the spread of misleading content in online environments.

Keywords: Fake News Detection, Machine Learning Classifiers, Fake News Dataset, Classifier Performance.

Abstract: The rapid growth of digital media has made the detection of fake news an essential task, as misinformation can quickly spread online and influence public opinion, decision-making, and social trust. This study explores the effectiveness of traditional machine learning techniques in classifying news articles as fake or real. Using the WELFake dataset, which contains 72,134 news articles from the Kaggle platform, classifiers such as Support Vector Machine (SVM), Random Forest, Decision Tree, and Gradient Boosting were evaluated. Initial experiments achieved strong results, with several models reaching an F1-score of 0.90. Further improvements were obtained by engineering additional features, leading to an enhanced F1-score of 0.96. The findings highlight the capability of traditional machine learning approaches for fake news detection and provide insights into building effective models to mitigate the spread of misinformation.

Keywords: Include at least 4 keywords or phrases.

Abstract: As artificial intelligence (AI) has spread to most industries, cloud environments have emerged as the foundation for deploying and expanding smart applications. While the compatibility of AI and cloud computing brings their synergy closer to perfection, their matching raises new security threats in the form of adversarial attacks, data exfiltration, and expanding attack surfaces. This research discusses existing threats, analyzes AI-powered security systems, and identifies the expanding utilization of machine learning for threat detection and autonomous response. A comparative evaluation of legacy security and AI-based security strategies identifies that legacy systems deliver basic defense while AI contributes maximally to resilience, precision, and responsiveness. Future enhancements such as autonomous AI agents, quantum-resistant cryptography, and real-time sharing of threat intelligence are also discussed with the objective of framing next-generation secure AI- cloud infrastructure.

Keywords: Artificial Intelligence, Cloud Computing, Threat Detection, Adversarial Attacks, Quantum-resistant Cryptography.

Abstract: International food security is largely dependent on farming, but crop diseases continue to threaten crop quality and yield. To minimize losses and maintain sustainable agriculture, plant diseases must be accurately and promptly diagnosed. Automation is very desirable because the traditional methods of visual inspection and laboratory analysis are frequently laborious, subjective, and unavailable in remote locations. Crop disease detection has been transformed by recent advances in deep learning (DL), which enable models to automatically extract discriminative features from plant photos without the need for human assistance. From manual scouting and preliminary image processing to conventional machine learning and more recent state-of-the-art deep architectures, this review tracks the development of disease detection techniques. Across popular crops like rice, maize, and tomatoes, pivotal techniques like convolutional neural networks, transfer learning, ensemble methods, and vision transformers are critically reviewed and compared. Examined are real-world uses like drone imaging, precision agriculture systems, mobile applications, and IoT-based monitoring. Along with fascinating potential directions like multimodal learning, cloud–edge AI fusion, and farmer-centric design, challenges like sparse datasets, environmental heterogeneity, computational cost, and unexplainability are discussed. This review provides a comprehensive picture of creating reliable, field-deployable crop disease detection systems by synthesizing improvements and shortcomings.

Keywords: Agriculture; Deep Learning; Crop Disease Detection; Precision Farming; Convolutional Neural Networks; Vision Transformers.

Abstract: Social media platforms such as Facebook, Instagram, Tik Tok and Twitter have evolved. into complex socio- technical structures which affect society, technology, and the world economy. Artificial intelligence (AI) sentiment analysis and recommender systems have made personalization and predictive analytics better. Technologically speaking, though such advances cause concern as well. Concerning algorithmic bias, privacy, surveillance and the spread of fake information. Socially, such platforms enable activism, learning and community building, especially noticeable during such periods of world crisis, but they also lead to issues such as disinformation, cyberbullying, digital addiction and ideological polarization. Economically, social networking sites are robust ecosystems that encourage entrepreneurship, influencer marketing, and digital commerce and are also associated with risks of fraud or monopolistic control and data commodification. This review offers an objective perspective of their dualistic character of both causes of ethical, social, and governance dilemmas and catalysts of innovation by synthesizing the findings of 35 peer-reviewed articles published between 2017 and 2025. These findings indicate the urgency of making the operation transparent, innovative strategy sustainable and responsible use of AI. The evolving landscape of the social networking platforms will require that accountability and digital literacy is enhanced to ensure that opportunities to be exploited and risks reduced.

Keywords: Networking Platforms, Artificial Intelligence, User Behavior, Digital Marketing, Misinformation, Data Privacy.

Abstract: Depression, affecting over 280 million individuals globally, imposes an economic burden exceeding $1 trillion annually through reduced productivity and healthcare costs. This paper presents an innovative hybrid system integrating natural language processing (NLP) and facial recognition to identify early depressive symptoms in students, utilizing the Reddit Self-reported Depression Diagnosis (RSDD) dataset and ethically sourced classroom imagery. Combining textual analysis (TF-IDF, BERT embeddings) with facial feature extraction (HOG, PCA, FaceNet), the system achieves 0.92 accuracy, 0.90 F1-score, and 0.94 AUC, surpassing NLP-only (0.90 accuracy, 0.88 F1-score, 0.91 AUC) and facial recognition-only (0.85 accuracy, 0.83 F1-score, 0.87 AUC) baselines. In a case study with 500 students, it identified 87% of at-risk individuals, demonstrating practical utility. The methodology employs robust preprocessing, feature fusion, and real-time processing tailored for educational settings, enabling efficient monitoring and intervention. Ethical safeguards, including differential privacy, data anonymization, and informed consent, address privacy concerns and mitigate biases in Reddit’s predominantly young, male demographic. Designed for scalability, it supports mental health interventions and attendance tracking, offering a cost-effective solution to promote student well-being. By integrating advanced machine learning with ethical frameworks, the system aligns with global mental health strategies, reducing the burden of undiagnosed depression. Its modular design enables adaptation to diverse educational contexts, highlighting the potential of multimodal approaches for complex mental health challenges.

Keywords: Depression Detection, NLP, Facial Recognition, Deep Learning, BERT, FaceNet, Mental Health, Ethics.

Abstract: In industries such as construction, manufacturing, and chemical processing, Personal Protective Equipment (PPE) plays a critical role in protecting workers from serious injuries. Even with strict safety rules in place, many workplaces struggle to ensure consistent PPE use, often due to negligence or lack of constant supervision. Relying on manual checks is time-consuming, error-prone, and impractical for large-scale monitoring. This study presents a real-time PPE detection system that combines computer vision with deep learning to address these challenges. The system uses the YOLOv8 object detection model to identify key safety items—helmets, safety vests, and face masks—directly from live video streams. A diverse and annotated dataset of industrial scenarios was used for training, enabling the model to reach a mean Average Precision (mAP) of 96%. The results show that the system can accurately and quickly detect PPE usage, offering a practical, scalable, and cost-effective alternative to manual oversight. By reducing reliance on human monitoring, this approach can improve compliance, enhance workplace safety, and help prevent avoidable accidents.

Keywords: PPE detection, YOLOv8, deep learning, computer vision, workplace safety, real-time monitoring.

Abstract: An Internet of Things (IoT)-driven quality control mechanism within the manufacturing sector commences with the visualization of item movements throughout production lines, aimed at identifying critical stations and pathways that have a substantial impact on product quality. This preliminary phase yields valuable insights into the production flow and highlights potential bottlenecks or opportunities for enhancement. Feature Engineering is employed to derive pertinent information through the selection and transformation of features, thereby augmenting the efficacy of machine learning models. The performance of the model is evaluated in comparison to other classification methodologies, such as Support Vector Machine, Naive Bayes, Random Forest, and Gradient Boosting, predicated on the chosen features. Through the examination of the interrelations among features, stations, lines, and the response variable, a deeper comprehension of the most influential factors affecting product quality and defect occurrence is attained. By leveraging visualizations of production line movements, feature importance rankings, and classifier performance metrics, this IoT-driven framework furnishes actionable insights for manufacturers to enhance product quality and mitigate defects.

Keywords: Internet of Things, Sensors, Machine Learning, Classification, Manufacturing.

Abstract: The global energy transition is being defined by the integration of hybrid renewable energy systems (HRES) and the creation of intelligent smart grids. While this shift toward sustainable energy is essential, it brings new challenges, including the intermittency of renewables and the limitations of traditional energy management systems (EMS). Modern IT solutions, specifically microservices-oriented architectures (MOA) and cloud-edge computing, offer a path to overcome these issues by improving modularity and enabling real-time decision-making. This paper links these two domains, connecting traditional EMS with microservices-based frameworks to enhance renewable energy management and grid resilience. We begin by outlining key challenges: a lack of modularity, rigid centralized structures, and, crucially, the absence of robust resilience testing. Drawing on a review of over thirty studies, we explore recent advancements in microservices, HRES optimization, and cloud-edge computing. We note a significant gap in the energy sector's use of chaos engineering and failure injection methods, which have proven successful in the IT world for validating system resilience. Our findings reveal that while HRES and cloud-edge technologies are studied individually, their combined application within a microservices-driven EMS framework is largely unexplored. This review highlights the potential of using microservices to containerize EMS functions and facilitate real-time cloud-edge collaboration. It also emphasizes the critical need for advanced resilience testing to ensure fault tolerance. Future research should focus on standardizing EMS service decomposition, integrating AI- and blockchain-based coordination, and, most importantly, adopting chaos engineering frameworks to validate the resilience of future smart grids. This will create more adaptable and reliable energy management systems for the modern power grid.

Keywords: Microservices, Smart Grid, Hybrid Renewable Energy Systems, Cloud-Edge Computing, Energy Management System, Resilience, Chaos Engineering

Abstract: This paper studies the dual role of AI as both a contributor to and a mitigator of climate impact, focusing on its capacity to reduce carbon footprints across various sectors. AI techniques, such as intelligent energy management, predictive maintenance, supply chain optimization, and smart transportation, enable more efficient use of resources and reduction of greenhouse gas emissions. Furthermore, AI-driven optimizations in its own development through model compression, green neural architectures, and carbon-aware scheduling are transforming how AI systems are built and deployed with minimal environmental impact[1]. By integrating AI into climate strategies and promoting energy-efficient AI development, the technology can serve as a powerful enabler of global decarbonization efforts. This paper highlights key innovations, challenges, and pathways toward leveraging AI for a more sustainable and carbon-conscious future.

Abstract: Human–Machine Interface (HMI) design is increasingly moving towards more intuitive, contactless, and efficient modes of interaction. This paper presents a gesture-based framework in which humans communicate with machines or online platforms via normal hand gestures. Integrating computer vision with machine learning, the system described can capture, recognize, and understand gestures in real time without the need for conventional input devices. All identified gestures are assigned a pre-determined command so that the interface is applicable in healthcare, industrial automation, and home automation, where physical contact can be restricted or unwanted. The system prioritizes accuracy and responsiveness to deliver smooth user experience and also provides major benefits for users with disabilities. Overall, the research presents gesture recognition as a reliable, hygienic, and user-friendly option for next-generation human–machine interaction. Keywords-Human–Machine Interfaces (HMI), Intelligent Interface, Computer Vision, Hand Gesture Control, Machine Learning, Real-Time Interaction, Touchless Interface, Image Processing, User-Centered Design, Technology for Accessibility, Natural User Interface (NUI), Human–Computer Interaction (HCI), Sensor-Based Interaction, Contactless Control.

Abstract: The abstract Artificial Intelligence (AI)-powered chatbots are some of the most creative applications of AI, which has significantly changed customer service. Customer service has traditionally relied on human operators to consistency respond to inquiries, resolve issues, and ensure customer satisfaction. However, the cost, scalability, and of current approaches are limited as the demand for multilingual communication, 24/7 support, and personalized experiences increases. A dependable solution to these issues is provided by chatbots that use artificial intelligence (AI) and are built on machine learning (ML), natural language processing (NLP), and deep learning models. This study looks at the planning, creation, and evaluation of an AI-powered chatbot for customer support. The paper begins by examining the challenges that customer service is now facing and outlining the justifications for integrating AI in this industry. It offers thorough design specifications, system requirements, and a feasibility assessment using UML diagrams and architectural modeling. The development process, which involves selecting appropriate lifecycle models, frameworks, and algorithms for chatbot implementation, is also covered in the project. White-box and black-box testing are two testing methods used to ensure the system is correct and effective.

Abstract: Cyberspace, cybercrime, cybersecurity, and cyberculture are only a few of the many ideas associated with digital technology that go under the umbrella word "cyber". With the development of AI, hackers are using ChatGPT and other similar technologies more frequently to carry out complex cybercrimes. Natural language processing-powered ChatGPT may produce polymorphic malware, believable emails, and false information, which makes it simpler for thieves to trick victims. Artificial intelligence is changing the world of cybercrime. Hackers are now using tools like ChatGPT to make their attacks more sophisticated. This AI can create highly realistic phishing emails and even malware that's hard for antivirus software to detect. This is a big problem because AI is essentially giving less-experienced criminals the ability to commit much more complex and serious crimes. With AI, a small-time crook can now launch attacks that were once only possible for skilled professionals. To stay safe, it's crucial to adopt strong cybersecurity habits. Make sure you're using good antivirus software, creating strong, unique passwords, and keeping a close eye on your financial accounts. If you do become a victim of cybercrime in India, remember that there are legal options available to help you. The rise of AI in cybercrime means that our defences must also evolve to keep up with these new, smarter threats.

Abstract: Fake news has emerged as a crucial challenge in the contemporary digital landscape, where misinformation proliferates rapidly across social media and online platforms [1][7][10]. The origins of this issue trace back to observations that false information disseminates more swiftly than factual content, leading to widespread societal disruptions, including distorted public perceptions and decision-making processes [2][9][12]. The core problem addressed in this studies is the inadequacy of conventional fake news detect-able methods, which are often labor-intensive and ill-equipped to manage the voluminous daily influx of online content [3][6][13]. Human-led manual verification proves inefficient for processing millions of articles, exacerbating the scalability issues in real-time environments [4][14]. To mitigate this, the propose-able framework settings for learning algorithms integrated with Natural Language Processing (NLP) techniques for automated classification of news articles as authentic or fabricated [5][8][15][18][22]. By extracting and analysing textual features, linguistic patterns, and stylistic elements—such as TF-IDF-based selections and sentiment analysis—the system enables rapid processing of vast datasets [6][9][10][19][23]. This hybrid approach, incorporating ensemble methods and other learning models like BERT, facilitates efficient detection and enhances accuracy across multilingual and cross-platform contexts [8][11][16][17][18][20][25]. Ultimately, the project endeavours to develop an intelligent system that safeguards users from deceptive content, fosters reliable information ecosystems, and upholds the integrity of news consumption in society [21][24].

Keywords: fake news detection, machine learning, NLP, text classification, supervised learning, feature extraction, social media analysis, information verification, automated detection, news authenticity

Abstract: Cardiovascular disease (CVD) is a major global health challenge, contributing significantly to morbidity and mortality. With the continuous rise in incidence rates, there is an urgent need for advanced analytical methods to assist in early detection and diagnosis. This study explores the application of data mining techniques on a Transthoracic Echocardiography Report dataset to predict heart disease. Using the Knowledge Discovery in Databases (KDD) methodology, nine iterative steps were applied to process and analyze 7,339 echocardiography reports collected from a hospital. Three predictive models—J48 Decision Tree, Naïve Bayes, and Neural Network—were developed and evaluated. Experimental results indicate that all models achieved strong predictive performance, with the J48 Decision Tree yielding the highest classification accuracy of 95.56% and superior True Positive Rate. These outcomes demonstrate the potential of data mining-based approaches in enhancing diagnostic reliability and supporting cardiologists in clinical decision-making.

Keywords: Cardiovascular disease, Echocardiography, Data mining, Knowledge Discovery in Databases (KDD), Predictive modeling, Decision Tree, Naïve Bayes, Neural Network.

Abstract: Artificial intelligence (AI) is quickly becoming an important presence across a spectrum of areas and perhaps has one of the greatest impact in health care. This paper presents a review of the literature related to the utilization of AI technologies in health care particularly in relation to diagnosis, treatment, patient monitoring, efficiency in work processes and ethical considerations. We conduct a scoping review of the literature, synthesizing the main findings relating to potential gain and barriers to implementation of AI in health care practice. The findings suggest a strong role for AI in enhancement of diagnostic accuracy, efficiency, and personalized treatment. Meanwhile challenges of equity and algorithmic bias, data privacy and regulation particularly will need to be overcome prior to more widespread AI integration into practice. In conclusion, there is opportunity for AI to change the landscape of health care, yet the sustainability of this will depend on sound governance, ethical frameworks and contributions to interprofessional practice across fields.

Keywords: Artificial Intelligence, Healthcare, Machine Learning, Medical Imaging, Predictive Analytics, Drug Discovery, Patient Monitoring, Ethics in AI.

Abstract: The COVID-19 pandemic has sparked a global increase in study into understanding and predicting the disease's transmission, intensity, and effects. Machine learning (ML) has emerged as a significant tool in this quest, allowing for the analysis of large and complicated datasets to identify patterns and generate accurate predictions. This literature review synthesizes information from 10-15 peer-reviewed publications and review articles that investigate the use of machine learning algorithms in COVID-19 prediction. The algorithms used in the examined studies include Support Vector Machines (SVM), Random Forests (RF), Artificial Neural Networks (ANN), Convolutional Neural Networks (CNN), and hybrid models. These models have been applied to many datasets, such as clinical records and imaging data. Epidemiological data are used to forecast infection rates, disease severity, hospitalization risk, and mortality. The abstract compares the performance of several models, emphasizing the advantages of ensemble and hybrid techniques for increasing prediction accuracy. Data quality, model interpretability, and generalizability are among the issues discussed. The analysis indicates that ML models, particularly those that combine several algorithms and data sources, have great potential for improving public health responses and decision-making during pandemics. Future research directions include the development of real-time predictive systems, integration with existing epidemiology models, and ethical considerations when applying machine learning in healthcare settings.

Abstract: The rising proliferation of fake news on digital platforms has become a serious social and technological challenge. Fake news, which is defined as false or misleading information that is passed off as reality, has the potential to sway elections, exacerbate divisiveness, and jeopardize public health. Existing detection techniques, which include deep learning, transformer-based architectures, and handmade machine learning classifiers, perform well in benchmark scenarios but have three enduring issues: lack of explainability, domain shift, and adversarial paraphrase. This study combines explainability modules, cross-domain evaluation, and adversarial data augmentation to present a strong NLP-driven framework for false news identification. The pipeline compares traditional models such as Support Vector Machines with newer designs like BiLSTM, BERT, and RoBERTa, coupled with a hybrid ensemble (BERT + SVM). Several benchmark datasets from the political, health, and entertainment domains—LIAR, FakeNewsNet, BuzzFeed, FEVER, Weibo, and Hinglish—were evaluated.The results show that hybrid ensembles achieve improved robustness against adversarial attacks and temporal drift, whereas transformer-based models continuously outperform previous methods. The results emphasize how crucial it is to integrate factual, stylistic, and semantic elements in order to create robust and interpretable fake news detection algorithms for practical uses.

Abstract: Educational institutions face significant challenges in managing course registration processes, including manual paperwork, long queues, registration errors, limited accessibility, and inefficient administrative workflows. Traditional course registration methods are time-consuming, prone to human error, and lack real-time updates on course availability. An Online Course Registration System provides a comprehensive solution by leveraging web technologies to create a streamlined, efficient, and userfriendly platform for students and administrators. This web-based application enables students to browse available courses, register for their preferred courses, manage their academic schedules, and track their enrollment status from anywhere at any time. The system utilizes PHP as the server-side scripting language and MySQL as the database management system to ensure robust data handling and secure transactions. By implementing features such as real-time course availability, automated enrollment verification, payment processing integration, and comprehensive reporting tools, the system significantly reduces administrative workload while improving the overall student experience. The Online Course Registration System not only eliminates geographical and temporal barriers but also provides valuable analytics for institutional planning and decision-making. With its modular architecture and scalable design, this system represents a significant advancement in educational technology, offering institutions a reliable and cost-effective solution for modernizing their registration processes.

Keywords: Online Registration, Web-based System, Educational Technology, Course Management, PHP, MySQL