MATTER: International Journal of Science and Technology

WEB ATTACK DETECTION COMPARATIVE ANALYSIS OF LSTM AND DNN-BASED DEFENSE MODELS WITH CORRESPONDENCE ANALYSIS

2025-08-06T05:47:02+00:00

Recently since there exists more companies using web, web attacks to hijacking or manipulate the privacy information have increased. Among web vulnerability OWASP has introduced, SQL injection, XSS, File Inclusion have constantly occurred through more than a decade. It concludes that web servers have trouble with blocking old-fashioned web vulnerabilities. This paper is going to skim through web attack defending methods and compares existing web attack detection machine learning models and new ensemble model DPL with ANOVA, chi-square analysis, correspondence analysis to find out relativity between model and web attack. As result of correspondence analysis, brand new model DPL excels existing models but even DPL model have low relativity on XSS. It is expected that post research must introduce more XSS relevant model.

EXPLAINABLE PREDICTIVE ANALYTICS FOR SMART HEALTHCARE USING A MODULAR HYBRID INTELLIGENCE FRAMEWORK

2025-09-09T04:58:05+00:00

The exponential growth of healthcare data – driven by electronic health records (EHRs), wearable sensors, and continuous remote monitoring systems – has created both immense opportunities and complex challenges for modern clinical decision-making. Effectively harnessing this heterogeneous, high-volume, and high-velocity data requires intelligent systems that can not only deliver accurate predictions but also provide interpretable insights to support clinician trust and patient safety. In this paper, a modular hybrid computational intelligence framework designed to advance personalized, real-time healthcare analytics, is proposed. Our approach synergistically integrates deep learning for high-dimensional feature extraction, fuzzy inference systems for transparent reasoning under uncertainty, and genetic algorithms for adaptive optimization. This tri-layered architecture enables the system to learn from multimodal data sources, including physiological signals (e.g., ECG, glucose levels), structured clinical records, and unstructured patient-reported outcomes, to predict critical health risks such as cardiac arrhythmias, myocardial infarction, and diabetes-related complications. Through experimentation on publicly available real-world datasets, the proposed framework demonstrates superior predictive accuracy, enhanced interpretability, and computational efficiency compared to conventional machine learning and deep learning baselines. Importantly, the inclusion of fuzzy logic modules allows clinicians to trace back the reasoning paths of the system, addressing the growing demand for Explainable AI (XAI) in regulated healthcare environments. This research bridges the longstanding gap between model performance and transparency by offering a scalable and modular solution that is adaptable to diverse clinical contexts. By supporting proactive risk stratification and timely interventions, the framework has the potential to transform reactive care models into intelligent, preventative, and patient-centric healthcare delivery systems.

EFFECT OF ADJACENT SHADING ON BUILDING ENVELOPE HEAT GAIN IN TROPICAL CLIMATE

2025-09-11T06:09:15+00:00

In tropical regions, approximately 60% of building energy is consumed by cooling systems, with heat gain through the building envelope being a major contributor. The Overall Thermal Transfer Value (OTTV) is a metric used to quantify average heat gain in air-conditioned buildings. However, the standard OTTV calculation does not account for shading from adjacent buildings—an increasingly common feature in high-density urban areas. This paper presents an empirical study on the thermal performance of building envelopes considering adjacent shading in tropical climates. Dynamic simulations of annual heat gain were conducted for buildings with and without adjacent shading for comparative analysis. The results highlight the significant impact of adjacent structures on heat gain performance and offer supplementary data to improve the current OTTV calculation method, especially for multi-block developments.