Evolutionary discriminant functions using genetic algorithms with variable-length chromosome

We propose a method of determining discriminant functions to improve the performance of pattern recognition. The discriminant function is a linear combination of functions that are a product of power of the input information. the proposed method consists of genetic algorithms and multiple regression analysis. Genetic algorithms with variable-length chromosome search forms of functions. Multiple regression analysis calculates the coefficients of terms. Experiments were performed for various tasks including an acoustic diagnosis for compressors as a real world task. The results showed that the proposed method was effective to improve the classification performance