| Summary: | 碩士 === 雲林科技大學 === 資訊管理系碩士班 === 98 === TKA (Total knee arthroplasty) are widely recognized as being successful and effective for the treatment of degenerative joint disease. From literature review, the related TKA studies mostly use statistical methods. These methods for themselves are fraught with complexity, uncertainty and non-linear problem in terms of medical datasets may be unable to more accurately finding important information. The medical datasets typically include a large number of features (attributes), some features are irrelevant, and therefore it can’t intuitively understand the corresponding to main factors which affecting the resource utilizations of healthcare.
In order to solve the problems mentioned above, this study employs specialist advice to filter relevant cases (records) and proposed an integrated five features selection methods to select the important features, furthermore, use rough set theory (RST) to extract the rules and compare with other methods in accuracy. The proposed model contains: (1) data screening by specialist opinion, (2) two stage feature selection by ANOVA and proposed an integrated feature selection approach, (3) data discretization and rule generation by RST. In verification, this study use three different sub-datasets for comparison accuracy, the first one is through the ANOVA test to obtain 35 features, the second is through the integrated five kinds of feature selection method to get three important features, and the third is through the opinions of specialists to obtain 5 features. After 10 times run repetitively, the average accuracy of proposed model is 68.82%, 92.54%, and 91.99%, respectively. The results can provide to NHI as reference, for setting the TKA standard in the future , such as in the TKA operation, there is a minimum experience threshold can effectively enhance the of healthcare resource utilization.
|
|---|
