Research On Personalized Recommendation Algorithm Based On Heterogeneous Information Network

In recent years,the rapid development of Internet technology has greatly enriched the content of the network.On the one hand,the abundant network information greatly facilitates People's daily life;on the other hand,the huge scale of information makes it difficult for people to get the information they need quickly.Under this background,recommendation system emerges as the times require.It can analyze users' interests and preferences by using their historical behavior information,and then generate recommendations that meet users' needs.At present,the research and application of recommendation system have made great progress.Some studies have shown that heterogeneous information network can be used to effectively analyze the historical behavior information of users and applied it to the recommendation system as an information modeling method,which can deeply explore the implicit feedback information of users.However,the existing recommendation algorithm based on heterogeneous information network still has the problems of low quality information and data sparsity,which will lead to the deviation of the prediction of user preference,and then reduce the recommendation effect.In view of the above problems,this paper considers to improve the recommendation algorithm from the perspective of filtering low-quality information and incorporating other auxiliary information.The specific research contents are as follows:(1)Aiming at the problem that low-quality information will affect the recommendation effect,this paper proposes an information filtering mechanism based on heterogeneous information network.Specifically,the rich semantic information in the heterogeneous information network can be obtained through the meta-path,and the information acquired in different meta-paths has different semantics and different information quality.High-quality semantic information can more accurately reflect the user's personalized information,which is conducive to the generation of more accurate recommendation results.However,the low quality semantic information not only interferes with the recommendation,but also increases the operation scale of the algorithm,thus reducing the recommendation effect.Therefore,this paper will evaluate the quality of semantic information from the perspective of meta-path,and screen out the low-quality semantic information,so as to alleviate the negative impact of low-quality information on recommendation.(2)This paper considers to integrate the user's label information into the recommendation,on the one hand,to alleviate the problem of data sparsity,on the other hand,to obtain the user's subjective attitude towards the item.Specifically,the meta-path can be used to mine the underlying semantic relations of entities in the heterogeneous information network,and then make recommendations by virtue of the potential relations between users and objects.However,it is impossible to know the real attitude of the user towards the item when recommending,while the user's label describes the user's understanding and evaluation of the item and reflects the user's subjective attitude towards the item.Therefore,the inclusion of the user's label information in the recommendation can not only alleviate the problem of data sparsity,but also more accurately predict the user's satisfaction with the recommended items,so as to achieve more accurate recommendation.(3)Based on the scenario of heterogeneous information network,this paper also proposes a personalized recommendation algorithm that integrates users' social information.The core content of this algorithm is divided into two parts:one is to analyze the potential relationship between users with the help of heterogeneous information network,and integrate explicit social relationship with implicit social relationship,so as to analyze the personalized characteristics of users more accurately;The second is to consider the influence of trust relationship and distrust relationship among users on the recommendation effect,and the mutual influence of trust relationship and distrust relationship,and to construct a regular term to be added into the matrix decomposition model to evaluate user similarity.In summary,based on the heterogeneous information network scenario,this paper proposes two improved personalized recommendation algorithms.One of them considers the improvement from two aspects of information filtering mechanism and the integration of user tag information,and the other considers the integration of user social information for improvement.Finally,the effectiveness of the proposed algorithm is verified by experiments on two real data sets.