A dynamic segmentation method of power customer based on rough clustering

This paper studies the issue of segmentation results change over time, on which existing research paid limited attention, and proposes a dynamic segmentation method for power customers based on rough clustering. With the selection of power consumption indicators, value analysis systems of power customer are built and the weight of each indicator is determined by combination weighting. Rough k-means algorithm is used to construct a classifier and a two-dimension clustering segmentation method of current and potential value is presented. Two indicators, named the relative size and change rate of roughness of clusters, are proposed and describe changes of segmentation results in different periods combined with the changing cluster memberships of individual customers. Finally, a power supply company is taken as an example to illustrate the process of the proposed method and verify its feasibility.