An improved fuzzy modeling algorithm. II. System identification

For pt.I see ibid. In this part we focus on the fuzzy system identification stage. The Fuzzy C-Means (FCM) clustering algorithm is investigated. Two indices are introduced for assignment of number of rules and level of "fuzziness". An efficient method is also suggested for choosing the initial location of cluster centers. The classification problem is addressed in fuzzy modeling; and an efficient strategy is proposed for assignment of dominant input variables and their membership functions.