Improving evolutionary training for Sugeno Fuzzy Inference Systems using a Mutable Rule Base

The accurate modeling of a time series using a Sugeno Fuzzy Inference System (FIS) requires an algorithm that can train the FIS to minimize the error of seen and unseen data points. Many researchers have used genetic algorithms to optimize the parameters of the FIS membership functions with a great deal of success. It is presented here that incorporating FIS structure identification into the training process can greatly improve accuracy of predicting future time series data, by using the well-known Mackey-Glass time series as a benchmark. The main structural identification consists of optimizing the number of membership functions per input and total number of rules in the rule base.