The Design And Implementation Of Personalized Recommender System Based On The Analysis Of Online Customer’s Recessive Behaviors

With the continuous development of internet technology and e-commerce,the number and variety of network goods and services become more and more.In the face of such a large scale,how to help users find the right goods and services quickly and efficiently become a problem to be solved.Most of the existing method of personalized recommendation collaborative filtering methods conformity recommend the most suitable goods to user from many other goods with the same or similar functions.However, in jiangsu telecom online business hall(hereinafter referred to as Net hall), a typical business scenarios, exist some problems.such as data sparsity, "cold start", and recommendation accuracy.In allusion to data sparsity,it builds social tagging for goods based on user’s explicit and implicit data,As the important basis for personalized recommendation.In allusion to "cold start",it classify the user with data mining methods.To a certain extent,it reflects the user’s personalized interest preference.In view of the existing personalized recommendation technology is applied to the actual problems existing in the jiangsu telecom online business hall,it does some research on personalized recommendation mechanism based on the recessive data of customer online.The main work includes the following several aspects.(1)This paper proposes a kind of method, based on social tagging with recessive behavior data of the user.Through the analysis of user browsing page, residence time and other recessive data such as the operation of the user behavior,In combination with explicit data such as transaction records and the service evaluation,it builds the social tagging model in order to depict the function of the commodity attribute and non-functional attributes. Also it provides evidence for the analysis of the personal preference of users for goods.(2)This paper proposes a personalized recommendation method based on social tagging model.According to user’s constructed set of labels it finds the target users of neighbors,And based on the commodity feature vector,it assembles all goods which the target users and their neighbours have bought. At last calculate target user’s preference in the goods according to user label of goods and recommended for target users to meet the individualized functional requirements set of goods.At the same time the use of clustering methods to improve the chance of a new release goods are recommended.At the same time,the use of clustering methods Increases the chances of new release goods are recommended.(3)This paper proposes a personalized recommendation method based on user classification.For the "cold start" problem among new registered users lack of implicit data,it uses data mining method based on bayesian model to classify the user and recommend related products for specific types of users.After collecting enough explicit and implicit behavior data,it adjusts the result of the recommendation.(4)With personalized recommendation algorithm refer to this article,it designs and implements personalized recommendation system for online customers to verify the feasibility and effectiveness of the theoretical research results in this paper.In conclusion,in this paper,it does some research on the problem of personalized recommendation,combines a social tagging model of explicit and implicit data based on online customer,proposes a personalized recommendation method based on social tagging model,solves the problem of "cold start" through classifying the user with bayesian model.Finally, Verifies the feasibility and effectiveness of the results in this paper through a series of simulation experiments and the prototype system implementation.