Detecting temporal changes in customer behavior

Extant research has studied customer behavior in a static manner. But customer clustering can be used to identify the dynamic behavioral patterns of customers over a period of time. We develop a method for extending the standard fuzzy c-means clustering algorithm for detection of temporal changes in customer data. The study using real-life data leads to detection of appearance of new clusters and disappearance of old clusters. Using cluster validity indexes the novel method is shown to lead to formation of clusters that are better than those produced by the fuzzy c-means (FCM) algorithm.