Automatic Generation of Membership Functions and Rules in a Fuzzy Logic System

Fuzzy logic is playing a significant role in many control and classification systems. This arises from its simplicity, natural language based construction, dealing with ambiguity, and its ability to model linear and non-linear complex systems. But, with larger number of input and output variables, the building process of fuzzy system manually becomes daunting and error-prone process. In this work we suggest new methods for automatically creation and generation of fuzzy membership functions and fuzzy rule base respectively. The membership functions are created adaptively from the training data set using histogram of each feature individually. And in the beginning, a rule is generated for each membership function, then according to the weight assigned to each rule and membership function, the membership functions are merged according to successiveness of their domain or support. Also, the rules are started to be co-operated and merged to enhance the classification process. The resulted system is flexible, and is able to receive more rules and/or membership functions if needed.