Research Of Mobile Electronic Commerce Recommendation Algorithm Based On User Behaviors

In recent years,with the rapid development of mobile internet and intelligent terminal technology,mobile e-commerce has become a new development direction of e-commerce.Due to the limitation of mobile display size,mobile e-commerce is facing with serious information overload.Therefore,it is urgent to provide personalized recommendation service for users.The existing recommendation algorithms mainly use rates or implicit feedback to establish user profile model,then updates the model regularly to provide recommendation.However,the user's context will change with the change of time and location,which leads to the rapid change of user interest.The requirements of real-time recommendation in the mobile environment can not be satisfied by regularly updating model.In addition,due to the fact that the new user has no behavioral data or historical data is very sparse,the existing algorithm can not provide reliable recommendation results for new users,which affects the quality of recommendation.Aiming at the above issues,we propose a personalized recommendation algorithm based on purchase intention and interest degree,and a collaborative filtering recommendation algorithm based on multi-source information fusion.The main contributions of this paper are summarized as follows.(1)Considering that interests of mobile users change rapidly,the existing algorithms can't response to users' need in real time.This paper proposes a personalized recommendation algorithm based on purchase intention and interest degree(PIID).The concept of purchase intention is proposed to quantify the user's immediate interest of the items which are recently clicked or collected before payment.In offline mode,extracting users behavior characteristics,Through using logistic regression model to train the regression coefficient,we establish purchase intention prediction model.In order to precisely locate items that the target user may purchase,purchase probability is given by the linear combination of purchase intention and interest degree.According to the purchase records,using probabilistic latent factor model to learn users interest degree by maximizing the probability of purchase.While online,the user's real-time behavior characteristics are put into the prediction model,then we get the purchase intention of users.Through combined it with the user's interest degree,the items which has higher purchase probability are recommended.Experiments on real dataset show that the PIID algorithm has higher accuracy and F1 than the existing algorithms,and can provide efficient real-time recommendation for users.(2)Aiming at the problem that existing algorithms can not provide reliable recommendations for new users,This paper proposes a collaborative filtering recommendation algorithm based on multi-source information fusion.Firstly,using location information to establish a location consumption graph,we find the nearest neighborhoods set by calculating the effect of position distance and the preferences of different users to the target user.This algorithm recommends items for new user according to the neighbors preference.Next,we use PIID algorithm to gain the personal interests.Finally,the algorithm uses the purchase probability to combine the impact of nearest neighborhood preference with personal interest.We recommend the items which has higher purchase probability.Experiments on real dataset show that this algorithm has higher accuracy and F1 than the existing algorithms,and can provide a reliable recommendation for new users.(3)In order to verify the feasibility of the proposed algorithms,we design and implement a prototype recommendation system of mobile e-commerce based on users behaviors.