Optimizing heart disease diagnosis: a reinforcement learning-based ensemble method

The identification of heart disease is important as a means of early detection to enhance the quality of treatment. This research presents a new approach that combines ensemble learning and reinforcement learning to improve heart disease prediction. To capture the complex patterns in the data, the model employs a variety of classifiers which include Decision Trees, Support Vector Machines, and Artificial Neural Networks. A learning automata agent is incorporated at an appropriate point to learn the weights for classifiers that are then adjusted to improve the generalization of the learning process. The performance of the proposed model is confirmed by various experiments on the Cleveland Heart Disease dataset with an accuracy of 94.55 %, sensitivity of 96.55 %, and specificity of 92.31 %. Additionally, the statistical analysis results with (p < 0.05) demonstrated the significance of the improvements achieved by our approach. The proposed hybrid approach has several advantages such as higher accuracy, improved robustness, generalized learning, and computationally efficient. Using accurate predictions at an early stage, this model can go a long way in helping enhance the patients’ lives and lowering mortality associated with cardiovascular diseases.