The Use of Feature Engineering and Hyperparameter Tuning for Machine Learning Accuracy Optimization: A Case Study on Heart Disease Prediction

Heart disease (Cardiovascular) illness presents a noteworthy public health issue and ranks among the primary factors contributing to mortality worldwide. The World Health Organization (WHO) reports that approximately 32% of worldwide fatalities are attributed to heart disease. Consequently, it becomes crucial to implement preventive measures that enable the prediction of heart disease risks, aiming to mitigate its occurrence and decrease associated mortality rates. Several technologies and methodologies have been utilized to forecast the likelihood of heart disease by leveraging patient data and existing risk factors. One such approach is Machine Learning, specifically the Supervised Learning Binary Classification Technique of distinguishing between individuals with or without heart disease. Within this framework, the objective is to predict an individual's probability of developing heart disease based on specific features. The selection of these features is grounded in the strongest correlations observed in the available data. The researchers have identified several highly correlated features, namely ST Slope Up, ST Slope Flat, Exercise Angina, Oldpeak, Chest Pain Type, Max HR, and Sex. The objective of this research is to create an advanced predictive model that enhances precision by employing Feature Engineering and Hyperparameter techniques in a specific case study centered around forecasting the likelihood of heart disease. The results are promising, with the initial stage before the hyperparameter tuning 69.57% (data validation and 68.12% (data testing). After that, the model achieved an accuracy of 82.61% (data validation) and 86.23% (data testing) with the aid of the K-Nearest Neighbors algorithm with Hyperparameter Tuning GridSearch.