Heart Attack Prediction Using Machine Learning Classification Models

Heart disease is a chronic disease that must be taken seriously. One type of heart disease is a heart attack. Diagnosing any of these diseases requires a large number of evaluation factors and is often a tedious process for paramedics. There are many factors that must be evaluated, but it can be simplified by looking for the factors that have the most influence on this disease using the correlation matrix and selectKBest. From these two techniques, it was found that chest pain type was the variable that most related of prediction heart attack using dataset in this study. In addition, machine learning can help patients and paramedics to diagnose a heart attack by performing a classification model. This study aims to evaluate classification models for predicting heart attacks including decision trees, logistic regression, naive bayes, k-nearest neighbor, XGBoost, support vector machine (SVM), and random forest. Evaluation result of this study shows that the SVM model obtains the highest accuracy, f1-score, recall, and precision value with 85,53%, 88,17%, 95,35%, 82% respectively. Meanwhile, to see the performance of each model using the Mean Cross-Validation Score, it was found that logistic regression was the best model with a value of 0.8239.