Modified dynamic fuzzy c-means clustering algorithm – Application in dynamic customer segmentation

Abstract

The dynamic customer segmentation (DCS) is a useful tool for managers in implementing marketing strategies by observing dynamic changes that are happening in the customer segments over time. The Crespo’s dynamic fuzzy c-means (CDFCM) is one of the clustering algorithms introduced in the literature for DCS. We have suggested modifications to the CDFCM algorithm owing to certain shortcomings found in it, resulting in the modified dynamic fuzzy c-means (MDFCM) algorithm. To show the performance of the MDFCM algorithm, extensive experiments were carried out in comparison with the CDFCM algorithm using a retail supermarket dataset with eleven new data updates. To validate the results of the MDFCM algorithm, the fuzzy clustering evaluation measures such as Xie-Beni (XB) index, within sum of squared error (WSSE), root mean squared error (RMSE), Kwon index, and Tang index are utilized. The experimental results show that MDFCM is the most effective clustering algorithm for DCS, and the results are tested statistically to show its significance. The MDFCM algorithm is further compared with another successful algorithm available in the literature called Fathabadi’s dynamic fuzzy c-means (FDFCM). To show the usefulness of the MDFCM algorithm, a DCS framework is proposed and it has been demonstrated through a case study.