The fuzziness parameter m is an extra parameter that facilitates the iterative formulas of Fuzzy c-means (FCM). However, the parameter m, commonly set to be 2.0, is an important factor that effects the effectiveness of FCM. In literatures, the statistical study of m is so far not available. Viewing m as a random variable, we propose a novel idea to optimize the fuzziness parameter m. For the model selection, a modified cluster validity index is defined as the optimal function of m and improve the effectiveness of FCM. Then the simulated annealing algorithm is applied to approximate its estimate.
Lecture Notes in Artificial Intelligence, Vol.LNAI, No.10313