Heart Disease Prediction Using Machine Learning Approach with Data Preprocessing Optimization

Abstract

Heart disease is one of the leading causes of death in the world. Hence, the development of early prediction systems based on technology has become an important need to support medical diagnosis. Several machine learning algorithms have been applied to heart disease prediction, but their performance is highly dependent on data quality, preprocessing, and the suitable selection of the algorithm. However, prior studies have focused more on the application of specific algorithms and have not studied in detail the comparative influence of data preprocessing and optimization on the performance of different classification models. This study proposes a comparative approach of five machine learning algorithms, namely Logistic Regression, Decision Tree, Random Forest, K-Nearest Neighbor (KNN), and Support Vector Machine (SVM), combined with data preprocessing through data cleaning, outlier handling using Interquartile Range (IQR), and Min-Max Scaling normalization. The dataset used contains 270 medical records of patients with heart disease, each with 13 clinical symptoms. The dataset was divided into 80% training data and 20% testing data for the model training and evaluation process. Evaluation metrics used to assess the performance of each algorithm include accuracy, precision, recall, and F1-score. By analyzing the resulting data, the most effective algorithm in predicting heart disease based on the given clinical attributes was determined. The evaluation was conducted using the accuracy, precision, recall, and F1-score matrices.The testing results indicate that the K-Nearest Neighbor (KNN) algorithm provides the best performance with an accuracy of 93%, followed by Logistic Regression at 91%, while the Decision Tree has the lowest performance. The research results also show that the preprocessing stage also plays a significant role in improving the stability of data and classification model performance, especially for algorithms that are sensitive to the scale of features and the distribution of data. This study shows that by integrating preprocessing and choosing the right algorithm, it can increase the accuracy of heart disease prediction and has the potential to support the development of AI-based clinical decision support systems.