| Summary: | 碩士 === 長庚大學 === 電機工程研究所 === 86 === In the thesis, we propose a fuzzy expert system with the learning scheme, which can be applied on medical expert system. The software realization on the prostate cancer stage prediction system demonstrates the effectiveness and precision of the fuzzy expert system with the learning scheme.
Firstly, the design methodology of fuzzy expert system with the learning scheme will be shown. We use the linguistic variables and fuzzy if-then rules to record and integrate the knowledge of experts, and build a fuzzy expert system. The inference process of fuzzy expert system can be represented as an analytic function. In addition, we use similarity relations to measure the difference between the inference results of the system and the reference data provided by human experts. Therefore the learning scheme can apply backpropagation theory in neural network to modify the parameters of the fuzzy expert system. By the scheme, we can solve the revision problem of fuzzy expert system. Hence the fuzzy inference engine can operate correctly with the variance of time and place.
Secondly, the consequent application on prostate cancer stage prediction will demonstrate the implementation process and the experiment results. We get realization materials and experiment data from the from the Dr. Herry P. L. Chang who is the chief of Department of Urology in Chang Gung Memorial Hospital. The fuzzy expert system and its learning shceme are constructed in C++ program. For purpose of applying the program to the real system easily, we design the program in a flexible configuration with Graphic User Interface. We use 60 patients* data to examine the fuzzy expert system, and the system after training.
In the future, the application of the fuzzy expert system and its learnig scheme in Chinese Medical System, which needs a large database will researched. We will propose a new modular fuzzy relational database, which contains traditional relational database as a special case.
|
|---|
