Comparing the Performance of Data Conversion Model with Heart Disease Data Classification Using Machine Learning

การเปรียบเทียบประสิทธิภาพของรูปแบบการแปลงข้อมูลร่วมกับการจำแนกข้อมูลโรคหัวใจโดยใช้การเรียนรู้ของเครื่อง

Abstract

          This research aimed to study 2 types of nominal to numeric for heart disease prediction data classification. (1)Unique integers, and (2)Dummy coding, together with 5 machine learning methods including naive bayes, neural network, decision tree, support vector machine(SVM), and k-nearest neighbor(K-NN). The heart disease dataset was retrieved from the kaggle online database, which contains 12 features and 918 observers. The efficiency of our data classification was evaluated by 10-fold cross-validation using rapidminer. The results showed that the Unique integers and Dummy coding approaches gave different results with an accuracy of ±1.00 percent. Unique integers with the SVM method provided the best data classification efficiency at 87.58%, followed by the neural network method at 85.73%. The accuracy of the naive bayes, decision tree, and K-NN methods were 84.42%, 84.09%, and 71.45% respectively.