Using Data Mining Techniques to Detect Hospital Fraud and Abuse on Claims for Diabetic Medical care

• Main Authors: Jean-yi Chen, 陳靜怡
• Other Authors: Fen-May Liou
• Format: Others
• Language: zh-TW
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/423j43

碩士 === 元培科學技術學院 === 經營管理研究所 === 96 === Abstract This research uses on the comprehensive National Health Insurance (NHI) Research Database offered by the Bureau of National Health Insurance (BNHI) of Taiwan. This paper collects longitudinal data from diabetic ambulatory service in 2003. Based on the diabetes-mellitus and insurance expenditure previous studies, the studies selected potential variables with reference to Logistic regression, which is ranked second on prediction accuracy among 32 classification algorithms (Lim, Loh and Shih, 2000). The assumed a major position in the healthcare field as a method for predicting or classifying health outcomes based on the specific characteristics of each individual case (Chae et al., 2001). Therefore, the research aims building a data-mining model by Logistic Regression, Neural Network and Classification Tree algorithm for detecting fraudulent healthcare claims and compared three algorithm difference form diabetes of Taiwan. The results shows fraud detection models have 100%, the whole accurate rate showed the prediction models established by Classification Tree model (99%) better than neural networks (96%) and then Logistic model (92%).