Optimizing Multilayer Perceptron Classifiers for Predictive Heart Disease Diagnosis: A Hyperparameter Tuning Approach

Heart diseases, collectively known as cardiovascular diseases (CVDs), remain one of the foremost health challenges across the globe, taking millions of lives each year. This study evaluates the effectiveness of Multilayer Perceptron (MLP) classifiers in predicting CVDs using the Heart Disease Dataset from Kaggle. We implemented a structured approach using data pre-processing, model training with baseline MLP classifiers, and hyperparameter optimization utilizing Random and Grid Search techniques. The optimized models, particularly the MLP Classifier trained with Grid Search, outperformed the baseline, achieving up to 98.54 % accuracy and 100% precision. However, analysis of learning curves indicated the need for further model refinement to enhance generalization on unseen data. This study underscores the importance of hyperparameter tuning in developing effective MLP classifiers for CVDs prediction and suggests directions for future research, including exploring advanced optimization methods and regularization techniques.