Development of a heart attack prediction method using random forest and bat algorithm

Cardiovascular disease (such as heart disease, cancer and stroke) is one of the biggest causes of death worldwide with a percentage reaching 32% of all global death cases. According to the World Health Organization (WHO), among the numerous instances of cardiovascular fatalities, around 85% are caused by heart attacks. Detection of a person's potential for a heart attack needs to be done as an effort to prevent and reduce the risk of death. Machine Learning (ML) approaches have great potential to address problems in the computational biomedical domain, such as heart attack prediction. The Decision Tree (DT) algorithm is often used for predicting diseases in the health sector such as Covid-19, Parkinson's, tuberculosis, heart disease and several cases related to cardiovascular disease (heart, cancer and stroke) because it has higher accuracy results than other ML algorithms. The use of several DT algorithms can be developed to improve and stabilize the built model, such an algorithm is called Random Forest (RF). RF has the advantage of more accurate and stable predictions. RF requires hyperparameter tuning or searching for optimum parameter values to produce an accurate predictive model. The Bat Algorithm (BA) heuristic method is used in this study as hyperparameter optimization in RF. This article aims to find the optimum value of the hyperparameter through BA used in RF to produce accurate predictions and calculate the performance of the model being built. The heart attack dataset is collected through secondary data on the Kaggle database. Experimental findings indicate that the RF model's accuracy without BA optimization only reaches 80-82%, while the accuracy of the RF model optimized by BA can reach 90-93%. It means that the proposed method in this article is much better than the RF model without hyperparameter optimization.