Research On Recommendation Algorithm Based On Attribute Network And Comment Emotion

Come in with internet,the information on the Internet is growing exponentially.How to find the content that users are interested in from the huge amount of information and make personalized recommendations to users has become an important research topic.Through in-depth research,researchers have proposed a variety of recommendation algorithms.among these recommendation algorithms,collaborative filtering recommendation algorithm has always attracted much attention in academia and industry,and improves the quality of recommendation service to a certain extent.however,the algorithm still has the following problems:1.The traditional collaborative filtering algorithm takes the item score matrix data as the only input,but in social media,there is a certain relationship between users and the users in their watch list,and make full use of the attention relationship between users.we can better mine the potential points of interest between users and obtain similar preference information,but the attention relationship between users and users can not be used effectively in the traditional collaborative filtering algorithm.2.The collaborative filtering algorithm only uses the user's scoring data for the project,and on social media,the user's comment on the project is also very important information,which contains the user's emotional response to the project.It can better express the user's preference than a simple score,but the traditional collaborative filtering algorithm does not use the comment information effectively.In order to solve the above problems,this paper makes use of user network relationships and user-centered comment information in social media.The main research contents of this paper are as follows:In order to make full use of user concerns and user attribute information this paper proposes a collaborative filtering recommendation algorithm based on attribute network representation.This algorithm uses the user concern network and the user's static and dynamic attribute information,uses the improved semi-random walk deepwalk algorithm to generate the user's low-dimensional vector representation in the user relation network,and calculates the similarity of the obtained user vector to generate the final recommendation result.Aiming at the effective use of comments,this paper proposes a recommendation algorithm based on SVM fine-grained emotion classifier.This algorithm models the Chinese vocabulary of massive corpus,selects seed emotion words,calculates the Euclidean distance between comments and seed emotion words,obtains the emotion type of comments,and uses the function based on interest attenuation to generate vector representation for each comment,and finally generates SVM emotion classifier.The final predicted score is used to modify or fill the user score matrix.On the basis of the above work,an EMSC(Embedding Multi Sentiment Combined)recommendation model is proposed,and on the basis of the recommendation model,a personalized movie recommendation prototype system is designed and implemented.In this paper,we use crawler technology to crawl Douban movie short review records and user attention list information,and compare this algorithm with the benchmark algorithm,show that the recommendation algorithms proposed in this paper increase the hit rate and NDCG index by an average of 2% and 1.6% respectively compared with the contrast algorithm,which can recommend movies more accurately.