An Efficient Deep Learning-based Model for Heart Diseases Prediction

Abstract

Heart disease remains one of the leading causes of mortality worldwide, necessitating effective predictive models for timely diagnosis and intervention. This study explores the application of deep learning techniques in predicting heart disease, leveraging the power of data mining and machine learning. Initially, electrocardiogram (ECG) numeric datasets are preprocessed to extract relevant features crucial for classification. Convolutional Neural Network (CNN) is employed as a classification technique due to its superior performance compared to traditional methods. Evaluation metrics including accuracy, precision, and F-measure are computed to assess the efficacy of the CNN model against the baseline K-Nearest Neighbors (KNN) classifier. Results indicate that CNN outperforms KNN, establishing its efficacy in heart disease diagnosis on the given dataset. Furthermore, a hybrid approach integrating logistic regression and neural networks is proposed for enhanced predictive accuracy. Logistic regression identifies significant risk factors contributing to heart disease based on statistical p-values, while irrelevant factors are pruned. The resultant significant factors serve as inputs to the neural network, which is trained to predict the likelihood of heart disease. This integrated approach demonstrates promising results in predicting heart disease, highlighting the potential of combining statistical analysis with deep learning techniques for improved diagnostic accuracy.