Abstract
The conventional fuzzy inference system is effective in accuracy, but needs a large number of parameters. Small number of input rule modules need small number of parameters, but they are not effective in accuracy. To overcome them, small number of input rule module with linear transformation has been proposed. If these models are evaluated by using AIC(Akaike's Information Criterion) or BIC(Bayesian Information Criterion), the model with high accuracy and small number of parameters is selected. In this paper, we will show that the small number of input rule module with linear transformation shows high accuracy and needs small number of parameters by using AIC and BIC, and determine the number of rules for the small number of input rule modules with linear transformation by using AIC and BIC.
