Rule induction of breast cancer medical diagnose based on combination of rough sets, artificial neutral network and genetic algorithm

Research on Computer-aided diagnosis (CAD) of breast cancer in the last decade is reported tremendous. In this paper, a breast cancer diagnosis system based on rough set theory, SOM neutral work, and Genetic Algorithm is proposed. With this system, we transform data to knowledge which is very important to early diagnosis and further medical research of breast cancer. The system first discretizes breast cancer medical data by training SOM neural network; then reduce condition attributes by GA; finally induce diagnosis rules from decision table. Prior knowledge is not required by system. Specific diagnosis rules are available, and result is objective. Testing system on WDBC data set, we get 48 breast cancer diagnosis rules. Use these rules as classifier to evaluate performance of diagnosis system. The experiment result shows that diagnosis accuracy is substantially increased compared with traditional diagnosis method and other CAD methods.