Research And Application Of Recommendation Algorithm Fusion Social Relations In E-commerce Environment

With the rapid development of social networks,accurately identifying and mining their corresponding social circle information from the user's huge virtual relationship network can not only greatly facilitate the management of the user's personal relationship network,but also plays a very important role in personalized recommendation,influence communication,and public opinion warning.Due to the different mechanisms of each network platform,the difficulty of discovering social circles in different platforms is also different.In the Weibo platform,the blog posts published by Weibo users and the user's attention and interaction behavior contain explicit and clear social information.Researchers can use this information and design algorithms to realize the discovery of explicit social circles,And combine it with the corresponding recommendation algorithm to serve personalized recommendations of Weibo users.However,although e-commerce platforms such as JD.com and Amazon contain available information that reflects user preferences such as user network behavior and user comments,they do not have clear social mechanisms and interactive information like Weibo.Social recommendation algorithms that use social information to improve recommendation quality cannot be directly applied to e-commerce platforms.To this end,based on the user's online behavior and product reviews in the e-commerce platform,this article extracts various implicit social information contained in it,designs a method to measure the implicit social relationship between users,and constructs an algorithm for discovering implicit social circles.It also combines the implicit social circle with a social recommendation algorithm based on matrix decomposition to improve the accuracy of the e-commerce platform recommending products to users.The specific work of this article includes the following three aspects:(1)A hidden social circle discovery method based on user behavior is proposed.In e-commerce platforms,users' online behaviors mainly include purchasing behaviors and scoring behaviors,and users tend to score according to their preference for commodities.Since the user's preference for the product will not change in a short time,from this point,this article looks for users with similar interest preferences according to the user's rating matrix R for the product,and gives a formal definition of the implicit social circle On the basis,the method of mining hidden social information between users is designed.(2)A hidden social circle discovery method based on user comment content is proposed.User reviews directly reflect a user's most genuine thoughts about a product in his heart.Therefore,when the user makes a purchase decision,the comment information of the product can be used as an important reference basis for predicting whether the user purchases the product.To this end,this article uses the topic model LDA to analyze the comments given by the user to obtain the topic preference vector of the user's corresponding comment,and then calculates the similarity of the topic preference vector between users,and selects the topic preference that is most similar according to the similarity of the topic Of the top-k users as the main members of the hidden social circle,thus achieving the construction of the hidden social circle.(3)It realizes the algorithm of product recommendation by combining the implicit social circle with the social recommendation algorithm based on matrix decomposition.Starting from two points of user behavior and user content,mining and discovering a user set with high similarity to the target user forms a hidden social circle.Finally,the extracted implicit social circle and the social recommendation algorithm are fused to obtain the user's rating prediction matrix for the product,and the top-k products with the highest predicted ratings are selected for effective product recommendation for the target user.