Research On Collaborative Filtering Algorithm Based On User Interest And Time

With the rapid development of the Internet and mobile Internet,there are more and more websites in the fields of movies and music,which leads to a sharp increase in information in the network.It takes a lot of time and effort for people to obtain the required information.The recommendation system is an effective method to solve the problem of "information overload".Among them,collaborative filtering is a widely used and successful algorithm in the recommendation system.However,with the increase of the number of users and the number of projects,the collaborative filtering algorithm faces problems such as data sparsity,difficulty in obtaining user interest preferences,and scalability.This thesis focuses on the first two issues and proposes three improved algorithms.The main research work of this thesis are as follows:1.Aiming at the problem that the user's interest preference is difficult to obtain effectively in the traditional collaborative filtering algorithm,a collaborative filtering algorithm based on user interest preference clustering is proposed.Considering that users have different preferences for each keyword,TF-IDF(term frequency–inverse document frequency)is used to calculate and establish user-keyword preference matrix.Based on Canopy algorithm and K-means,users are clustered and recommended.The experimental results show that the method can alleviate the data sparsity problem and reduce the time complexity of the recommendation algorithm while capturing the user's interest preference.2.Aiming at the problem that the traditional collaborative filtering algorithm does not fully consider the drift of user interest,a collaborative filtering algorithm based on adaptive time weight is proposed.The Pearson similarity calculation method is improved by traversing the undirected graph to screen out the user score data.In the calculation of the prediction score,the adaptive time weight function is introduced,and the time decay attenuation factor is used to control the decay speed of the time function,which can alleviate the problem of poor performance of the recommendation algorithm caused by data sparsity and user interest drift to some extent.3.Aiming at the cold start problem of recommended system,a collaborative filtering algorithm based on user features and project keywords is proposed.Based on the first two algorithms studied,the feature attribute-keyword correlation matrix and user-keyword preference matrix are established.The similarity calculation strategy is improved based on user content similarity and user preference similarity.The Logistic time function is used to fit the user's interest changes in the scoring prediction.The simulation results show the effectiveness and feasibility of the method.