OptiDeepNet: Optimized Deep Network for Efficient Prediction of Heart Disease

Abstract

The accurate Coronary Heart Disease (CHD) prediction is challenging due to the variety of risk factors and the high diagnostic costs. Several deep learning models have been developed to predict the CHD efficiently. However, the crucial influence of choosing optimal hyper parameters of deep learning model makes it difficult to predict CHD with any degree of accuracy. To address this challenge, this study introduces OptiDeepNet (Optimized Deep Network for Efficient Prediction of Heart Disease), an optimized deep network for efficient prediction of  heart disease that integrates  Random Forest for feature selection and Enhanced Long Short-Term Memory (ELSTM) for CHD prediction, thereby improving the predictive performance of the system. Specifically, feature selection is conducted using a Random Forest classifier to identify the most informative characteristics features which are significant for predicting CHD. Furthermore, the ELSTM model was proposed where the hyper parameters are optimized utilizing Advanced Cuckoo Search, a meta heuristic optimization algorithm. This optimization step aims to maximize the performance of a model to learn intricate temporal relationships in the Cleveland Heart Disease dataset, thereby enriching the model's overall performance. Extensive experimentation shows that OptiDeepNet achieves a predictive accuracy of 98.3%, providing a promising tool for reliable cardiovascular diagnostics.