The Research Of Book Recommendation Algorithm Based On User Comments

The rapid development of the Internet,especially the emergence of Web 2.0,provides large and rich information resources.While people swim in the ocean of information,"information overload" bring more and more confusion to people.Faced with a lot of information,people often don't know how to choose the information that they really need.Choosing information also need spending a lot of time and effort.Recommendation system appeared,which solved the problem of information overload,and has been widely used and become an important part of the electronic commerce platform.At present,collaborative filtering is one of the most widely used techniques,especially in the application of e-commerce platform.The traditional collaborative filtering recommendation result is based on the user's score.On the one hand,User-Item rating matrix is facing severe data sparsity problem.On the other hand,user's score reflects the user's overall preferences of purchase products,but user's preferences for product attributes can not be reflected from the overall score.In order to fully understand the user's preferences for different features of the product,a large number of researchers began to study how to mine user's opinion from the user comments,so as to provide users with more accurate recommendation.In this paper,the book recommendation algorithm based on user comments,mainly from the following aspects of the in-depth study and discussion.First of all,preprocess the user comments,extract the feature-emotional words,the different levels of the product features scores,and then get user's preference for different features of the product.Secondly,in the process of the item similarity score prediction,to use the Item-based on the score similarity and feature similarity to predict the score,to fill the User-Item rating matrix and solve the problem of data sparsity.Then in view of the problem that the traditional User-based collaborative filtering only uses the score information and user's preference,this thesis proposes the method of adding preference similarity calculation when calculating similarity of users.Finally this thesis validates that the algorithm is effective through the experiments,the experimental data from Stanford SNAP.The results of experiments demonstrate that our method could achieve a good recommendation effect compared with the traditional algorithm.