Study On Electronic Commerce Market Based On Customer Behavior Analysis

With the rapid development of the Internet, it becomes more prevalent that users go shopping through e-Commerce web sites which prompt the perpetual innovation of service providers to meet customers'requirements. As a kind of important information which highly reflects users'intent, the behavior of customers on web sites has significant research value. Through analyzing and mining them, service providers can obtain the interest and purchase intent of users so as to build customer interest model and customer demand model. Based on those models, they can change marketing strategies as well as adjust services on purpose to satisfy customers'experiences and enhance customers'stickiness. To further explore the e-Commerce market, this thesis comprehensively utilizes the statistical analysis, the feature analysis, and the multi-relational clustering analysis to analyze customer behaviors. Thus, we can help companies to well understand user intent and to promote innovative services in the e-Commerce market. In summary, the thesis includes the following works:1) In order to analyze the statistical characteristics of customer behaviors, we introduce methods from human dynamics to study statistics driven by customers at the population level, analyze the customer purchasing patterns, and compare the characteristics of products among different product fields and the similarities and differences of review behavior patterns. Through empirical analysis we obtained some useful results which provide basis for further investigation on feature analysis and clustering analysis.2) In view of the feature importance evaluation problems based on customer behaviors, we analyze the behaviors on new product customer with the decision tree induction. We have extracted customer behavior features, customer attributes, product attributes, and make full use of them to construct decision tree in order to distinguish the importance among features, and have further obtained useful behavior patterns.3) With the multi-relational problems which exist widely in e-Commerce data, the market segmentation research problems based on unified similarity calculation techniques have been investigated in this thesis. We introduce the similarity calculating method which is based on a unified relationship matrix. With the integration of complex relationships, we can indicate customers or products in homogenous space. In this way, the traditional data mining algorithms for clustering analysis can be applied to implement the segmentation of customers or products.Good research ideas and programs in e-Commerce application fields can be supplied if we conduct above works, such as the formulation of marketing strategies, technical support of recommendation systems. More than that, our works can offer appropriate suggestions for applications in other related fields.