| | 講 題 | Machine Learning for Diagnosis of Coronary Artery Disease | | | 講 者 | 國立中央大學 Yu-Lin Hsu, Chia-Ru Chung, Jorng-Tzong Horng | | | 日 期 | 2019/09/25 | 長 度 | 00:13:30 | 人 氣 | 101 次 | | | The main global cause of death is coronary artery disease by the report of the World Health Organization in several years. Furthermore, the medical costs of coronary artery disease are pretty high. The most importantly, heart disease is greatly killed Taiwanese in recent years. Thus, in order to reduce the harm to people, it is necessary to predict coronary artery disease accurately and earlier. The major purpose of this study is to construct different machine learning models on diagnosing of coronary artery disease. The Z-Alizadeh Sani dataset from UCI Machine Learning Repository, including 303 patients and 54 features, was mainly adopted in this study. We apply the 3-fold and 5-fold cross-validation and evaluate the accuracy, sensitivity, specificity, precision, Area Under Curve and Matthews correlation coefficient for different model algorisms, including decision tree, logistic regression and ensemble learning technique. The highly accuracy of 10 features model and 20 features model are 87.77% and 88.10%, respectively. It shows almost same outcomes of prediction in both models. Besides, Decision Tree model are 7% less than Logistic Regression one when predicting accuracy in the dataset containing 20 features. Also, the specificity of top 5 related features model is less than 50 percent. It proves that the significance of the top 5 related features. Only taking the 10 features from people and using uncomplicated machine learning models, the prediction outcome is good (The best Area Under Curve: 0.9322). And, to improve the accuracy of a single model, ensemble learning needs more features. Finally, researchers can use linear classifiers to predict coronary artery disease. Hence, diagnosing coronary artery disease will be easy and effective, so it can reduce the harm to Taiwanese.
| | | 提 供 | TANET台灣網際網路研討會-TANET2019 | | |
|
