An intelligent system employing an enhanced fuzzy c-means clustering model: Application in the case of forest fires

Abstract

Fuzzy c-means is a well-established clustering algorithm. According to this approach instead of having each data point Dp i = ( X, Y) belonging only to a specific cluster in a crisp manner, each Dp i belongs to all of the determined clusters with a different degree of membership. In this way cluster overlapping is allowed. This research effort enhances the fuzzy c-means model in an intelligent manner, employing a flexible fuzzy termination criterion. The enhanced fuzzy c-means clustering algorithm performs several iterations before the proper centers of the clusters “more or less” stabilize, which means that their coordinates remain “almost equal” to the previous ones. In this way the algorithm is expanded to perform in a more flexible and human like intelligent way, avoiding the chance of infinite loops and the performance of unnecessary iterations. A corresponding software system has been developed in C++ programming language applying the extended model. The system has been applied for the clustering of the Greek forest departments according to their forest fire risk. Two risk factors were taken into consideration, namely the number of forest fires and the annual burned forested areas. The design and the development of the innovative model-system and the results of its application are presented and discussed in this research paper.