Research On Personalized Recommendation System Based On Context Attribute Information

With the rapid growth of network information resources,it is very difficult for users to obtain the information quickly and accurately.The emergence of the search engine to solve the problem of a part of the user query information,but the search engine can not achieve the user's needs for personalized recommendation function.Personalized recommendation system is to achieve this function.Personalized recommendation system is based on the user's needs.For example,businesses can tap the user's preference information based on massive data,the potential customers to dig out and further expand the scope of sales,which has more consumer groups.The majority of the needs of users,personalized recommendation system can provide personalized recommendation to users according to personal preference characteristics of the user,the user can quickly select the goods they need in a large amount of information,thereby reducing unnecessary selection in the process of the choice of goods in time.Therefore,personalized recommendation system for both users and businesses are both practical and valuable.From the current situation of the development of recommender systems,it is widely known that collaborative filtering recommendation algorithm and content-based recommendation algorithm.Collaborative filtering recommendation algorithm is to analyze information of user behavior,the user will find out the target user behavior information of similar,similar according to user preferences on items to measure the target user preferences on items,the target user preferences for goods according to the degree from high to low ranking,according to the ranking results recommended list.Finally,the results are fed back to the target user.The main principle of content-based recommendation algorithm is: according to the characteristics of the user and the analysis of the characteristics of the project.The commonly used methods include: Bayesian model,neural network model and space vector model.The characteristics of the user is based on user preferences of the project information analysis.In the field of personalized recommendation system,these commonly used recommendation algorithm can effectively improve the conversion rate from the user's identity to the identity of the buyer,thereby enhancing the ability to sell.Although these commonly used recommendation algorithms have made great achievements,but only according to a single score data mining similar users and items,the recommended results are not verysatisfactory.At present,many scholars have added some contextual attribute information,such as labels,locations,and so on,to improve the effect of personalized recommendation.Based on reading a lot of literature,the key technologies of recommendation algorithm are studied,based on existing technology for innovative improvement recommendations of their own,and through the simulation experiment proves the feasibility and advantage of the scheme.The results of this paper are as follows:(1)the project context information which is evaluated by the user and the user context information which is evaluated by the user are integrated into the recommendation algorithm,which effectively improves the accuracy of the recommendation effect;(2)a personalized recommendation algorithm(CATD)based on context aware and tensor decomposition is proposed,and the simulation results on the Movielens dataset show that the algorithm is effective(3)using kernel density estimation technology and user context attribute information,project,project and user preference model were constructed,and put forward a new method for the calculation of similarity based on preference model,then the similarity value of high neighbor fusion nodes;In the end,we recommend users and projects with a certain recommendation method.