Research On Customer Consumption Behavior Based On Data Mining

With the rapid development of the world economy,and the spread of the internet industry,consumer consumption behavior has undergone enormous changes.Many enterprises begin to convert their core business model from the “product-centric” to the “customer-centric” in order to be in a dominant position in the competitive market economy,and pay more attention to the customer quantity.At present,there is relatively few and single content research on the customer consumption behavior.With customer consumption behavior being more and more diversified and hierarchical,the traditional method of analyzing customer behavior has not been able to adapt to the change of it.By analyzing the advantages and disadvantages of related algorithms in thesis,K-MEANS clustering algorithm is used to segment the customer,and the customer value matrix is set up by combining segment results with customer consumption behavior variables RFM.By analyzing the change of consumption behavior of different value customers in different periods,and combining with association rules,the customer consumption behavior can be forecasted.These related algorithms are analyzed and improved to get a better performance.The main work of thesis is divided into the following points:(1)An in-depth analysis is performed on the characteristics of customer consumption behavior.Thesis introduces the basic process of customer behavior research and compares the related research algorithm.Aiming at the problems in the existing methods,thesis presents a method that performs a cluster process before the customer behavior analysis with the association rule combines with the changing customer consumption behavior patterns.(2)The customer segmentation is performed according to the characteristics of customer consumption behavior such as diversity and instability.The traditional K-MEANS algorithm has some problems in accuracy and execution time for clustering analysis.Aiming at this disadvantage,a validity function is proposed in thesis to determine the optimal clustering number K,and the stable initial clustering center is obtained by the greedy algorithm and the distance balance function between classes.The modified K-MEANS algorithm is used tosubdivide the customer,and the effectiveness of the improved clustering algorithm is verified by off-line experiment.(3)The changes of customer consumption behavior patterns are analyzed to establish and evaluate the forecasting model.In order to solve the problem that the traditional forecasting model can only analyze the single attribute of the rule result in different period,thesis redesigns the method to calculate the similarity and abruptness of the rule in different periods,and increases the matching degree of attributes in the multi-attribute of the rule result part,and combined with the association rules to establish the forecasting model,which predicts the specific changes of consumer consumption behavior patterns in the future.Finally,the validity of the model is evaluated by comparing the average relative error and root mean square error of the model before and after the improvement.