Coronary Heart Disease Prediction from Common Risk Factors

Abstract

Coronary heart disease (CHD) is the most prevalent type of heart diseases, and is one of the most common chronic diseases leading to death. This study analyzed the data collected from coronary heart disease patients and developed a predictive model for coronary heart disease diagnosis. The dataset from Kaggle website contains 462 observations, one target variable and 9 common risk factors, such as systolic blood pressure (sbl), the use of tobacco each year, low density lipoprotein (ldl), body adiposity index (BAI), family history, the score of type A personality, body mass index (BMI), daily use of alcoho and age. At first, the data prepossessing indicates that there are no missing value and extreme distribution, and the correlation and collinearity between the variables are weak. Then the study develops two prediction models using the methods such as k-nearest neighbor (KNN) and logistic regression. To choose the most suitable one, the indicators like kappa value and accuracy are the main factors to evaluate the model’s performance. The results shows that the logistic regression model have the higher accuracy, which is equal to 0.7174. After dropping some variables which have little impact on final prediction, the outputs present that yearly use of tobacco, low density lipoprotein, family history and BMI are the main factors which may influence the diagnosis of CHD. In the end, more clinical patients’ data still needs to be collected for improving the model’s accuracy.