Fuzzy inference systems by genetic algorithm and factor analysis modeling for multivariate complex systems

The authors propose a system which can automatically learn causal relation for multivariate complex problems by use of fuzzy inference and genetic algorithm. It has been difficult to infer the correct results from a lot of input variables by using only the fuzzy inference. We first concentrate many variables into a few variables of the input of fuzzy inference by factor analysis. Secondly, the genetic algorithm and delta rule are used to adjust and learn the fuzzy inference rules. We apply this system to human behavioral system with many input variables. By this causal modeling, we can identify the complex human system more precisely than the regression analysis generally used. >