Enhancing Heart Disease Prediction and Classification using Machine Learning and Deep Learning Techniques

Abstract

Cardiovascular diseases (CVDs) remain the leading cause of death globally, high- newlinelighting the urgent need for effective early-stage detection. Accurately diagnosing heart newlinedisease at its earliest stages presents a significant challenge for medical professionals. newlineFortunately, advancements in modern diagnostic technologies offer promising solutions for timely identification and intervention. newlineThis research aims to connect the potential of Machine Learning (ML) and Deep newlineLearning (DL) algorithms to analyze different datasets related to cardiovascular health. newlineThe primary objective is to enhance the accuracy and reliability of predictive modeling for heart disease (HD). The research contributes to advancing predictive and classification analytics in healthcare, specifically focusing on cardiovascular conditions. To achieve this, methods are proposed such as applying Explainable Artificial Intelligence (XAI) techniques, a Soft Voting Ensemble (SVE) ML Technique, and a hybrid DL technique (RNN+GRU) for HD prediction. Polynomial Jacobian Matrix-based Deep Jordan Recurrent Neural Network (PJM-DJRNN) is used to predict and classify HD. newlineThe research employs various ML classifiers, including Random Forest (RF), Logistic newlineRegression (LR), Naive Bayes (NB), K-Nearest Neighbors (KNN), Gradient Boosting newline(GB), LightGBM (LGB), and AdaBoost (AB).To Enhance model accuracy, grid search newlinecv hyperparameter tuning technique,5-fold,10-fold cross-validations, and Soft Voting newlineEnsemble (SVE) techniques are used for prediction. A novel approach, the Polynomial newlineJacobian Matrix-based Deep Jordan Recurrent Neural Network (PJM-DJRNN) with newlineadvanced signal processing techniques, is presented for HD classification. newlineSVE achieved an accuracy of 94.1% (Cleveland dataset) and 95.85% (IEEE Data- newlineport dataset), surpassing individual classifiers. The hybrid DL technique (RNN+GRU) exhibited an accuracy of 93% on the IEEE Dataport dataset, emphasizing the im- newlineportance of data pre-processing in achieving this result. The PJM-DJRNN approach newlineachieved an impress