PREDICTION

Cardiovascular Diseases (CVDs) remain the paramount global health challenge, responsible for approximately 17.9 million deaths annually, with cardiac arrest representing a particularly critical and sudden event. Despite advancements in machine learning, accurate prediction is often hindered by highdimensional clinical data ("the curse of dimensionality") and the performance limitations of standalone classifiers. Consequently, this study aimed to develop a framework that integrates a novel Hybrid Recursive Feature Elimination-Firefly (RFE-Firefly) feature selection technique with a stacked ensemble classifier to enhance predictive accuracy. The study utilized a structured heart disease dataset from the Kaggle repository, subjecting it to rigorous pre-processing, including normalization and class balancing via the Synthetic Minority Over-sampling Technique (SMOTE). The proposed Hybrid RFEFirefly algorithm was implemented to optimize the feature space, successfully reducing the data to a parsimonious subset of seven critical features. For classification, a stacking ensemble model was constructed using five diverse base learners which are Decision Tree, Support Vector Machine, Logistic Regression, Random Forest, and XGBoost and unified by Random Forest as meta-learner to synthesize their predictive strengths. Experimental results demonstrated that the proposed Hybrid RFE-FA Stack Ensemble achieved a superior accuracy of 90.57% and an F1-score of 0.7619. This performance significantly outperformed individual base classifiers as well as ensembles using standalone feature selection methods. The study concludes that integrating hybrid feature selection strategy with an advance stack ensemble classifier effectively mitigates overfitting and provides a robust, highperformance tool for clinical decision-making in cardiac arrest detection.

The growing sophistication of phishing attacks continues to threaten digital security, causing financial losses, data compromise, and reduced trust in online platforms. Traditional blacklist and rule-based systems struggle to detect newly emerging phishing patterns, highlighting the need for adaptive, intelligent, and automated approaches. This study addresses this challenge by developing a machine learning-driven phishing vulnerability prediction model capable of accurately distinguishing phishing websites from legitimate ones. The objectives include dataset acquisition, feature preprocessing, model development, and performance evaluation. A labelled URL-based phishing dataset sourced from Kaggle was cleaned, encoded, normalized, and split using an 80:20 train-test ratio. Four supervised learning algorithms: Logistic Regression, Decision Tree, Random Forest, and Extreme Gradient Boosting (XGBoost), were implemented and assessed using Accuracy, Precision, Recall, and F1-score. Among all models, XGBoost achieved the best performance with an Accuracy of 0.96, Precision of 0.97, Recall of 0.96, and an F1-score of 0.96, demonstrating strong predictive ability and robustness. Random Forest also performed well but with slightly lower scores. The final XGBoost model was deployed in a Flask-based web application to provide realtime phishing detection and user awareness. The findings confirm that machine learning, particularly ensemble-based methods, provides a scalable and highly effective approach to phishing vulnerability prediction. However, limitations include reliance on static URL features, dataset-specific bias, and the absence of behavioural or real-time network indicators. Future work should incorporate hybrid features, adversarial robustness testing, and explainable AI techniques to improve transparency and adaptability to evolving phishing attacks.

This study focuses on evaluating the relationship between destructive and nondestructive testing methods for predicting the in-situ compressive strength of concrete. The main aim is to develop a reliable regression-based model capable of estimating concrete strength using non-destructive approaches. The study specifically examined Grade 20 (C20) and Grade 25 (C25) concrete to determine the correlation between rebound hammer results, and conventional compressive strength tests. The experimental procedure involved casting and curing concrete cubes and beams in the laboratory under controlled conditions. Both destructive tests (compressive and flexural strength) and non-destructive tests (rebound hammer) were carried out at curing ages of 7, 14, and 28 days, following BS EN 12390-3:2019, ASTM C39, and BS EN 12504- 2:2012 standards. Rebound hammer readings were taken before compression tests on each specimen to establish a correlation between rebound number and actual strength. The data obtained were analyzed statistically using regression techniques to develop predictive models capable of estimating compressive strength from non-destructive test results. The findings revealed that compressive and flexural strengths increased consistently with curing age for both concrete grades. At 28 days, C20 achieved an average compressive strength of 20.16 N/mm², while C25 reached 25.15 N/mm², aligning with their design targets. Rebound hammer values showed a strong positive correlation with destructive test results, with a prediction accuracy of about ±5%. The study concludes that properly calibrated non-destructive methods, particularly the rebound hammer tests, can effectively predict the in-situ compressive strength of concrete. This approach provides a cost-effective, rapid, and non-invasive means for quality control and structural assessment in modern construction practice.