Diagnostic Prediction Models for Cardiovascular Disease Risk using Data Mining Techniques

Abstract

Cardiovascular disease is the top national health problem that leads to a large number of deaths in Thailand. There is still a growing number of patients with the disease. Proactive measures of disease prevention and disease control are searching for risk groups. Therefore, people who are at risk can diagnose and manage themselves to reduce risk factors and adjust their behavior accordingly. For this reason, the idea of diagnostic prediction models for cardiovascular disease was conceived. The data of patients from 126 health promoting hospitals and 12 hospitals in Saraburi Province was collected. Then analysis was done to establish 6 models: logistic regression, random forest, back-propagation neural network, decision tree, naïve-bayes and K-nearest neighbors. Moreover, 10-fold cross validation was applied to each model. The results revealed that the logistic regression model achieved the highest accuracy rate, 99.940%, followed by the back-propagation neural network model, 98.105%.