Firstly, the advantages of subtractive clustering and fuzzy ??-means clustering (FCM) are combined to
automatically determine the initial number of clusters in FCM. Then, an improved close function which is used in the
function of validity to determine the cluster structure of FCM is proposed. The improved close function of cluster validity
index ignores the data that have a faint impact on the result of clusters and leads to a stronger adaptability and a faster
calculating. The experiment results show that the improved close function is effective and faster for computing.