Prediction of M-Commerce User Behavior by a Weighted Periodical Pattern Mining

Abstract: The rapid advance of wireless communication technology M-Commerce is not only being widely accepted but also it is being more used as a popular way of business / commerce done by portable devices. It is becoming an interesting to find patterns and prediction of mobile user behaviors such as their location and purchase transactions in mobile commerce effectively to provide the service. In this paper, it provides a more efficient service to the mobile commerce users by applying weighted frequent pattern and periodical pattern for prediction of purchase behavior of mobile users. The Mobile commerce Explorer consists of five major components: 1) Similarity inference model 2) Mobile Commerce Behavior Predictor (MCBP) 3) Weighted Mobile Commerce Behavior Predictor (WMCBP) 4) Weighted Mobile Commerce Behavior Periodical Predictor (WMCBPP) 5) Performance Evaluation. In a weighted frequent pattern method, by applying unique weights for each of the itemset and find the closest pattern along with support value. In addition, temporal periodical pattern method is used to find the frequent user behavior in all time intervals of the transaction including the weight of the each item set and support value of the user for an item. Finally, the percentage of precision and recall is measured by comparing the various methods to prove the efficiency of the proposed pattern mining and prediction.

Keywords: M-Commerce, User behavior, Similarity, Pattern, Prediction.