MLP-GA-RF:Hybrid Machine Learning with Fog Integration for Advanced Heart Disease Prediction

Abstract

Unawareness of improper detection and diagnosis are related to disease of heart being the main reason of mortality world-wide. This study overcomes the challenge proposed by this by putting forward a hybrid machine learning model, MLP-GA-RF, which is constituted of MLP (Multilayer Perceptron), RF (Random Forest) and Genetic Algorithm (GA). With high GA efficiency in optimizing the MLP parameters, this results in better predictive performance for this MLP, as well as providing a robust classifier for final diagnosis using RF. Finally, the system is integrated with a fog computing framework for further improvement of real-time diagnostic ability. This reason, among others, is why fog computing is so beneficial and differs from other forms of computing: it processes data nearer to the source, thereby lowering latency and decreasing reliance on centralized cloud infrastructure. This is a highly suitable decentralized approach for mobile applications related to healthcare such as monitoring of remote patient, and has the advantage of fast data analysis due to the distributed system. The recall, F1 score and AUC are rigorously evaluated on the model against conventional classifiers: Logistic, Support Vector Machines, Regression, XGBoost, Decision Trees, and Gradient Boosting. It is found that the MLP-GA-RF model outperforms baseline models across all substrates and consistently provides both more accurate and reliable predictions. Fog computing integrated with an optimized hybrid model is a great enhancement for prediction of heart disease.