Fuzzy rules extraction using self-organising neural network and association rules

Fuzzy logic is becoming popular in dealing with data analysis problems that are normally handled by statistical approaches or ANNs. The major limitation is the difficulty in building the fuzzy rules from a given set of input-output data. This paper proposed a technique to extract fuzzy rules directly from input-output pairs. It uses a self-organising neural network and association rules to construct the fuzzy rule base. The self-organising neural network is first used to classify the output data by realising the probability distribution of the output space. Association rules are then used to find the relationships between the input space and the output classification, which are subsequently converted to fuzzy rules. This technique is fast and efficient. The results of an illustrative example show that the fuzzy rules extracted are promising and useful for domain experts.