Journal of Data Mining and Management

SVM-RBF based Maternal Healthcare Model

2026-03-24T05:42:21+00:00

This paper presents a modeling based on Support Vector Machine (SVM) and Radial Basis Function (RBF) for the outcomes of maternal healthcare. This study proposes a machine learning–based predictive framework for maternal risk classification using a support vector machine with a radial basis function kernel. Maternal and child healthcare monitoring plays a critical role in reducing preventable morbidity and mortality. A structured dataset containing 50 clinical and demographic attributes was processed through systematic feature selection, normalization, and stratified sampling. The model was evaluated using a 70:30 train-test split and 5-fold cross-validation. Experimental results demonstrate an accuracy of 87%, with strong sensitivity and ROC performance. The predictive model was deployed via a web-based interface using Streamlit, enabling real-time maternal health risk assessment. The findings confirm the applicability of kernel-based learning in healthcare analytics. Future work includes ensemble modeling and explainable AI integration to improve interpretability and generalizability. The nutritional and hematological condition of individuals is indicated through records of Iron and Folic Acid (IFA) supplementation as well as hemoglobin levels, both of which are essential factors in determining maternal anemia and the overall risks associated with pregnancy. Furthermore, the dataset includes structured indicators of high-risk situations to capture significant clinical warning signs noted during prenatal evaluations. The outcome variable is defined as a binary class, indicating whether high-risk pregnancy status is present or absent. This organized blend of demographic, clinical, and preventive healthcare factors facilitates the creation of a strong predictive model for classifying maternal risk while ensuring a thorough representation of features for supervised learning analysis. The selected feature scaling was executed using standardization to normalize the distribution of numerical variables. Because Support Vector Machines are affected by the magnitude of features, standardization ensured that each attribute contributed equally to the formation of the decision boundary, subsequently enhancing convergence stability and predictive accuracy.

Mitigating Big Data Overload in IoT Ecosystems: A Comparative Study of Edge, Fog, and Cloud-Based Processing Architectures

2026-02-10T11:59:52+00:00

The explosive growth of Internet of Things (IoT) devices has led to the generation of massive volumes of real-time data, creating significant challenges in terms of storage, bandwidth, and processing efficiency, commonly referred to as big data overload. Traditional cloud-centric architectures struggle to cope with this influx due to increased latency, network congestion, and limited real-time responsiveness. This study investigates and compares the effectiveness of edge, fog, and cloud-based processing architectures in mitigating big data overload within IoT ecosystems. Through a simulation-based experimental framework, they evaluate each architecture against key performance metrics such as latency, throughput, bandwidth utilization, and energy consumption under varying data loads. Preliminary findings suggest that decentralized approaches, particularly edge and fog computing, offer superior responsiveness and bandwidth savings, while cloud architectures provide better scalability for long-term analytics. The study concludes with a proposed hybrid architecture that leverages the strengths of all three layers, aiming to optimize data flow and processing across diverse IoT applications. This work contributes to the design of more scalable, efficient, and resilient IoT infrastructures.

POSEatSea: A Hybrid Sentinel-1 SAR and AIS-Based Framework for Predictive Oil Spill Risk Detection at Sea

2026-03-30T11:38:16+00:00

Oil spill monitoring at sea is traditionally performed using remote sensing systems that detect surface contamination after an event has already occurred. To move from reactive detection to proactive risk assessment, this paper proposes POSEatSea, a hybrid framework that combines Sentinel-1 SAR imagery from Copernicus with AIS-based vessel behavior analysis. Sentinel-1 provides all-weather, day-and-night radar imagery for detecting spill-like ocean surface anomalies, while AIS data helps identify suspicious vessel movement patterns such as route deviations, unusual loitering, and irregular motion that may indicate elevated spill risk. The proposed workflow includes satellite image preprocessing, AIS cleaning, feature extraction, and machine learning-based classification to distinguish normal and risky maritime conditions. By fusing both data sources, the framework improves situational awareness and strengthens early warning capability for maritime authorities. The model is designed to support faster intervention, reduce environmental damage, and enhance monitoring of high-risk shipping corridors.

An Integrated Dealership Management System for Spare Parts, Jobcards, and Financial Analytics Using NoSQL Technologies

2026-03-17T06:16:35+00:00

With the tremendous growth in the number of car models and increasing service requirements, the automotive dealership business is in severe trouble in managing inventory levels for spares, managing customer information, job cards, and financial information. Such manual methods, either in hard copy or partially computerized, lead to inventory outages, incorrect invoicing, delayed reporting, and ineffective decision making. In this study, a Smart Automotive Dealership Inventory and Jobcard Management System was created for an automotive dealership business. The system combines serviceable parts management, accident parts management, and customer jobcard management into one online application. The advantage of having all components in one application allows the user to easily access and utilize inventory information, customer information, sales information, inventory notifications, and daily profit-loss statements. The front-end development of the application was completed with the following technologies: HTML, CSS, JavaScript, Tailwind CSS, and Bootstrap to develop a responsive dashboard design for the application. The back end, on the other hand, uses MongoDB, which manages different types of databases, enhances performance with indexes and aggregations, and ensures smooth operations, financial transparency, and the ability to make intelligent decisions. MongoDB is a database system that stores data in document form, and it is capable of managing different types of data, ranging from catalogs of spare parts and services to financial transactions. Using aggregations, the system is able to meet the needs of advanced reporting, providing management with the necessary information on sales, inventory turnover, and profits. Such a technology platform provides for high business efficiency, financial transparency, and decision-making for a dealership to gain long-term competitive advantage in a dynamic automotive service environment.

Hybrid CNN–Random Forest Framework for Real-Time Credit Card Fraud Detection

2026-04-02T08:53:21+00:00

Modern financial ecosystems increasingly depend on digital channels for everyday transactions, creating an environment where automated fraud detection has become operationally essential. High-throughput payment networks — spanning mobile wallets, contactless point-of-sale terminals, and instant fund transfer protocols — produce transaction streams at a scale that renders human-supervised screening impractical. Conventional rule-based screening mechanisms, though widely deployed, carry two well-documented structural weaknesses: their inability to respond to previously unseen fraud patterns without manual rule revision, and their tendency to misclassify valid transactions as suspicious, imposing both direct operational costs and customer-experience penalties. Overcoming these weaknesses simultaneously demands a machine learning solution capable of learning adaptive decision boundaries, scaling to high-volume transaction environments, and producing outputs that comply with emerging algorithmic accountability standards in regulated financial markets. To address these challenges, this paper proposes a two-stage sequential pipeline that couples CNN-based spatial feature extraction with Random Forest ensemble classification. During preprocessing, log transformation is applied to the heavily right-skewed Amount feature; all 30 input dimensions are rescaled to a [0, 1] interval using Min-Max normalisation fitted solely on training data; and SMOTE is employed exclusively within the training fold to counteract severe class imbalance (0.172% fraud prevalence) without contaminating held-out evaluation sets. Each processed feature vector is then restructured into a 5×6 two-dimensional matrix, allowing convolutional kernels to capture cross-feature spatial dependencies undetectable by flat-vector classifiers. The 384-dimensional abstract representation produced by the CNN is subsequently passed to a 500-tree Random Forest for final binary classification. Evaluation under stratified 10-fold cross-validation on the publicly available ULB Credit Card Fraud Detection benchmark demonstrates that the proposed hybrid model attains 98.6% classification accuracy, precision of 0.97, recall of 0.95, F1-score of 0.94, and AUC-ROC of 0.99 — surpassing five independently evaluated baselines (Logistic Regression, K-Nearest Neighbours, Support Vector Machine, standalone CNN, and standalone Random Forest) on every reported metric. Compared to the standalone CNN, the hybrid framework reduces false-positive alerts by 12%, while an average inference time of 4.2 milliseconds per transaction confirms viability for live payment authorisation systems. The proposed framework delivers a reproducible, interpretable, and rigorously validated contribution to the intersection of deep learning, data mining, and financial security.