Classification by reduction of the input attributes of medical systems with soft computing techniques

Detecting dominant inputs is an important problem in multi-input medical systems. In this thesis, a hybrid reduction system software based on a genetic algorithm and the rough set is developed to reduce the number of inputs and detect the dominant inputs in multi-input medical systems. The system performance is improved by using the proposed modified genetic selection mechanisms. The system was tested on both urological and standard databases and the dominant input values were calculated. Modifications added to genetic algorithm selection mechanism allow the software to support databases with high number of input variables. Local optimum problem of genetic algorithm based systems has been solved and effective memory usages have been obtained during tests.