Research on customer segmentation based on OCA cluster ensemble in customer crisis management

At present, single cluster model is adopted usually for customer segmentation. In this paper, clustering ensemble technology is applied to customer segmentation improves the cluster performance in customer crisis management owing to the uncertainty and fuzziness of customer behavior. First, this paper regards objective cluster analysis ( OCA ) as basic clusterer, which overcomes the shortcoming that current traditional customer segmentation methods cannot determine the number of clusters objectively, and then use a co-occasion matrix to combine the multiple clusterings to obtain the final consensus clustering. Finally, experiments show that the objective clustering ensemble algorithm based on random projection not only improves the accuracy of customer segmentation, but also finds out the optional clustering scheme in an automatic and objective way.