Automatic Fuzzy Rule Base Generation from Numerical Data: Cuckoo Search Algorithm

Fuzzy Classifiers are an powerful class of fuzzy systems. Evolving fuzzy classifiers from numerical data has assumed lot of remarks in the recent past. This paper proposes a method of evolving fuzzy classifiers using a three step technique. In the first step, a modified Fuzzy C–Means Clustering technique is applied to generate membership functions. In the next step, rule base are generated using cuckoo search algorithm. The third step was used to reduce the size of the generated rule base. By this method rule explosion issue was successfully tackled. The proposed method was applied using MATLAB. The approach was tested on a very well known multi dimensional classification data sets i.e. Iris Data. The performance of the proposed method was very encouraging. Further the algorithm is implemented on a Mamdani type control model for a battery charger data set. This integrated approach was able to evolve model quickly. General Terms Nature-inspired approach, Cuckoo Search Algorithm, Fuzzy Logic, Fuzzy Classifier, Rule Base.