Research On Collaborative Recommendation Based On Similar User And Sentiment Analysis

With the rapid development of the internet,the life is full of a lot of information,it is difficult to find out the specific content of the large data,this "information overload" phenomenon has become increasingly prominent and urgent need to solve.In this paper,we propose a collaborative recommendation algorithm based on similar user sentiment analysis,and personalized recommendation based on similar user interests.The user does not need to describe personal information needs,when the user has no definite demand it particularly.As an important technology to deal with information overload,personalized recommendation has been widely used in e-commerce,social networking and other fields.Collaborative filtering is the most successful and widely used technology in personalized recommendation because of its good recommendation effect and simple algorithm.Through the acquisition of the user's historical data,analysis and prediction of the user's potential interest,which recommend items.However,there are some problems in the application of the collaborative filtering algorithm in the application of the film review data,such as the serious spoken language,the sparse user behavior matrix,and the characteristics of the film project.To solve these problems,this paper launches the following research:First,according to the serious critics of colloquial,this paper proposes an extensible part of speech path algorithm,which can extract the emotion words in the text and avoid the influence of non-emotional words caused by non emotional words.An analysis of the part of speech path of training film critics,and sort out the generalization of part of speech.The word path generalization processing,reduced to prototype and then follow-up operation.It can simplify the complexity of the algorithm and reduce the number of parts of speech path.Build the lexical path library and sort by frequency of occurrence.Select a reasonable part of speech path manually,and increase the accuracy of part of the path.Second,the similarity calculation is the core of this paper.in order to solve the problem of sparse user scoring matrix,an improved Jaccard coefficient is proposed,The Jaccard coefficient is not accurate,and the similarity is not only related to the similarity of the viewing behavior,but also t the number of times that the behavior is consistent with the observation.Compared with the Jaccard coefficients through the experiment,it is proved that the coefficients proposed in this paper are more realistic and the MAE value is reduced effectively,and the recommended accuracy is higher.Third,in view of the particularity of the film review data,users are more interested in newer movies,so the probability of being seen is large,so this paper adds the time decay coefficient on the basis of the traditional recommendation algorithm.Comparing the traditional recommendation algorithm with the recommended algorithm of adding the critics data,it is shown that the recommended algorithm of adding the time attenuation coefficient is higher than the traditional method in the recommended accuracy and satisfaction.