Sparse Fuzzy System Generation By Cluster Estimation: A Projection Based Approach.
Alex Chong, Kok Wai Wong, T.D. Gedeon, László Tamás Kóczy · International Conference on Intelligent Information Processing · 2003
A projection based method for sparse fuzzy system generation is proposed. Given a set of training data, clustering is first performed on the output space. Data points from each output cluster are projected back to each input dimension forming one-dimensional clusters. The clusters from different dimension are then merged to form fuzzy rules. The number of rules to be generated can be controlled via a threshold parameter. Experiments have confirmed the effectiveness of the proposed technique. It is observed that the propose technique creates a sparse fuzzy system from data that achieves a satisfactory accuracy with reasonably few rules. The use of fuzzy rule interpolation further improves the accuracy.