Customer segmentation on mobile online behavior

Nowadays customers are becoming more and more personalized when enjoying various mobile online services, which highlights the importance of customer segmentation based on customers´ online behavior records. In this study, we propose a segmentation method that 1) divides the customer behavior sequences into cycles; 2) characterizes customers´ cyclical behaviors based on the probability density distributions from the temporal dimension and frequency dimension, which could investigate the customer behaviors more comprehensively; 3) calculate the customer similarity by computing the difference of the distributions; 4) adopts the k-medoid clustering algorithm to classify the customers based on the similarity matrix. The segmentation method is applied to a mobile online behavior dataset. The results of experiments indicate the relationship and typical customer usage regulations among various mobile e-commerce service groups, which will be informative for enterprises to understand their customers and improve their service quality.