Early Heart Disease Prediction Using Data Mining Techniques

Abstract

This study develops a predictive model for early heart disease detection using data mining techniques to enhance timely and accurate diagnosis. Heart disease prediction is complex due to the need to analyze various risk factors, such as age, cholesterol, and blood pressure. The model integrates multiple machines learning algorithms, including Random Forest, Support Vector Machine, and a hybrid ensemble approach, aiming to achieve higher prediction accuracy and reliability. The methodology follows five phases which include data collection, data pre-processing, feature extraction, model construction, and model evaluation. Data was gathered from publicly available health repositories, preprocessed to remove missing values and irrelevant information, and subjected to feature extraction techniques to identify influential predictors. The data was split into an 80:20 ratio for model training and testing to assess model performance across various classification algorithms. The hybrid model achieved an accuracy of 97.56%, precision of 98.04%, and recall of 97.09%, surpassing the individual algorithms tested. These findings indicate that the hybrid approach effectively supports early intervention for heart disease, particularly in healthcare settings with limited diagnostic resources. The study demonstrates that advanced data mining techniques offer a viable solution for improving patient outcomes through early detection of heart disease.