Heart Attack Risk Prediction with Duke Treadmill Score with Symptoms using Data Mining

Abstract

The healthcare industry has a huge volume of patients’ health records but the discovery of hidden information using data mining techniques is missing. Data mining and its algorithm can help in this situation. This study aims to discover the hidden pattern from symptoms to detect early Stress Echocardiography before using Exercise Tolerance Test (ETT). During this study, raw ETT data of 776 patients are obtained from private heart clinic “The Heart Center Bahawalpur”, Bahawalpur, South Punjab, Pakistan. Duke treadmill score (DTS) is an output of ETT which classifies a patient’s heart is working normally or abnormally. In this work multiple machine learning algorithms like Support Vector Machine (SVM), Logistic Regression (LR), J.48, and Random Forest (RF) are used to classify patients’ hearts working normally or not using general information about a patient like a gender, age, body surface area (BSA), body mass index (BMI), blood pressure (BP) Systolic, BP Diastolic, etc. along with risk factors information like Diabetes Mellitus, Family History, Hypertension, Obesity, Old Age, Post-Menopausal, Smoker, Chest Pain and Shortness Of Breath (SOB). During this study, it is observed that the best accuracy of 85.16% is achieved using the Logistic Regression algorithm using the split percentage of 60-40.