| Summary: | 碩士 === 國立雲林科技大學 === 工業工程與管理研究所碩士班 === 100 === The National Health Insurance (NHI) has been implemented in Taiwan since March 1995. In order to control the increasing medical expense, the Bureau of NHI started to implement the Diagnostic Related Groups/Prospective Payment System (DRG/PPS) in 2004. The accuracy of medical records is very important, because the disease classification staffs of hospital have to classify patients’ diagnosis records codes from medical records, especially the accuracy of coding complication / co- morbidity (CC). The CC classification not only affects the accuracy of DRG coding, but also has impacts on the interests of the hospital. The objective of this research is to develop an expert system to help disease classification staffs to make decision in the complication / co morbidity using several data mining techniques.
The accuracy results show that that decision tree based classifiers are better than the BPN based classifier. The C5.0 decision tree is better than the other decision trees in large sample size. On the contrary, the CHAID decision tree is better in small sample size. In additions, the Apriori algorithm, considering both minimum support and minimum confidence precision rate of 100% for the case of CC, truly can raise the overall accuracy rate.
|
|---|
