Using classification techniques to build mortality risk prediction model for patients undergoing hemodialysis

• Main Authors: Chung-Lun Yang, 楊仲倫
• Other Authors: Tzu-Ya Weng
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/6mvqxu

碩士 === 元智大學 === 資訊工程學系 === 104 === Background: Chronic kidney disease (CKD) has always been a highly prevalent disease in Taiwan. When the disease progresses to end stage renal failure (ESRD), the patient has be undergo long-term dialysis treatment. The cost of dialysis is immense and puts a huge burden on the health care system. The focus of this research is to examine clinical features associated with the progression of CKD to construct a model that can monitor the mortality risk of patients undergoing dialysis. Method:Using four classifiers (Decision Tree, Naïve Bayes, KNN, Random Forest) and three feature selectors (F-score,InfoGainEval, CfsSubsetEval), 6606 patients' dialysis data collected over the year 2007 to 2011 were analyzed for the building of a training model, and tested with a set of testing data to evaluate the performance of the constructed model. A website was also built to demonstrate the usability of the model. Results and discussion:The prediction model built with the decision tree classifier provided the best performance for predicting a patient's risk of mortality. The model has the potential to assist physicians and patients with the monitoring of their treatment response and disease progression.