Optimization Based Approach for Heart Disease Classification

Globally, heart disease is one of the main causes of death. Clinical data analysis is a huge problem when it comes to accurately predicting cardiovascular disease. This work presents a prediction model that makes use of numerous proven classification algorithms and different combinations of information. The goal of this work is to help in the detection of heart disease by employing a hybrid classification system depending on the Binary Harris hawks algorithm (BHHO) and the Logistic regression approach. Also, the Boruta algorithm with random forest is used and compared with the proposed PCA-BHHO algorithm. In this work, the data is first preprocessed, and missing values are filled with mean values. Then, data is scaled using standard scaler, and the proposed hybrid PCA and BHHO are applied to select the best features. RF and logistic regression are employed to classify the patients as heart disease patients or not. For comparison, Boruta is used for feature selection and RF for classification and compared the results with the proposed PCA-BHHO algorithm. Two datasets are utilized to test the proposed model: Statlog and the Cleveland heart disease datasets. The proposed PCA-BHHO algorithm attained an accuracy of 92.59% and 89.33% on the Statlog and the Cleveland datasets, respectively. At the same time, the Boruta-RF algorithm attained an accuracy of 90.14% and 87.64% on the Statlog and Cleveland datasets, respectively.