With the popularization and development of the Internet,there has been a trend of massive growth in different types of network data.Although the large-scale data volume meets the user's information acquisition needs to a certain extent,it also brings the "information overload" dilemma.In order to alleviate this dilemma,recommendation systems have emerged at the historic moment.Among them,recommendation systems with information filtering and data mining functions can find resources that meet the needs of target users from massive network resources according to the historical preferences of target users,and then recommend the resources.In the recommendation system,collaborative filtering recommendation algorithm is currently the most widely used recommendation algorithm.This algorithm is commonly used in e-commerce,movie recommendation and other fields,but the traditional collaborative filtering recommendation algorithm only considers the user's rating record of the item,determines the nearest user based on the similarity of the ratings,and then recommends the resources of the nearest neighbor to the target user.In the recommendation process,due to the sparseness of the scoring matrix,the quality of the selected nearest neighbor users is not high.Secondly,the algorithm also does not take into account the contextual factors and preference characteristics of the target user and the attribute characteristics of the project.Therefore,during the recommendation process of traditional collaborative filtering recommendation algorithms,although it can handle resources of various data types,the recommendation results are often unsatisfactory.In view of the above problems,this paper proposes a personalized recommendation algorithm that combines user portraits and user social relationships in specific situations.The main content of this article is as follows:(1)Aiming at the problem that the traditional collaborative filtering algorithm ignores the situation of the target user and their own preferences,this article introduces the user portrait technology based on the historical product characteristic data that the user likes and the online review data of the product,in order to effectively extract the target Features of user preferences and vectorization of user preferences using a space vector model.Secondly,in order to more accurately characterize user preferences,this article introduces context awareness technology and uses information entropy theory to analyze the impact of different context factors on users' choice of products.In order to determine the user's user preferences in a specific context;finally,the user portrait and context awareness are integrated into a collaborative filtering algorithm to form a recommendation algorithm that combines the user portrait and collaborative filtering in a specific context.(2)Aiming at the problem that the quality of the selection of the nearest neighbor users of the traditional collaborative filtering algorithm is not high,this article starts from the perspective of the user's social relationship and mines the ranking of the neighbor users associated with the target user.First,construct a directed graph of networks between users based on the association between users,and use the confidence of the association rules as the weight of the directed edges in the graph,and build a transition matrix based on this weighted directed graph.Second,calculate the user's The actual weight in social relationships defines the user with the higher weight as the nearest user of the target user.Finally,combined with the user portrait in a specific situation,a personalized recommendation algorithm that combines the user portrait in a specific situation and the user's social relationship is formed.More comprehensively solve the problem of "information overload" for users.(3)Empirical research.In order to verify the personalized recommendation algorithm that fuses user portraits and user social relationships proposed in this paper,taking real-world movie recommendations as research objects,the traditional userbased collaborative filtering algorithm,the fusion of user context factors and user portraits are analyzed separately.The comparison of recommendation algorithms and recommendation algorithms that fuse user portraits and user social relationships in specific situations.Experimental results show that the proposed algorithm has obvious advantages in recall rate,accuracy rate,and F1 comprehensive value.This research provides a reference for personalized recommendation,and also provides a two-way way for users and enterprises to obtain information,which has certain practical application value. |