Research On Collaborative Filtering Algorithm Based On Clustering And User Attributes

With the increasing popularity of the Internet,people have entered the era of big data.The amount of information on the Internet is like a sea,but there is very little information that is really useful to us.In the information resource explosion environment,if there is no good information filtering mechanism,people will be overwhelmed by the "ocean",and it will become more difficult to find the information they need.Therefore,in this context,the recommendation system is put forward.The recommendation system connects users with information and realizes the function of obtaining meaningful information for users from massive data.Collaborative filtering,as a mainstream recommendation algorithm,can analyze users' preferences from their history.However,there is a large data gap in the recommender system due to the small number of items that users participate in,which affects the recommender system's performance.In this paper,the technical aspects and relevant knowledge are introduced in detail.In view of the problems of low coverage and low precision caused by the recommendation of collaborative filtering algorithm,a series of related studies and activities are carried out to improve the collaborative filtering algorithm.The main research work is as follows:(1)Among the current personalized recommendation algorithms,the classic one is the collaborative filtering recommendation algorithm,which is the theoretical basis of many recommendation algorithms.However,there are still some problems affecting the recommendation effect.In this paper,the data sparsity of the data set was studied.The unrated missing values in the improved Slope One prediction scoring data were used to fill the predicted values into the original matrix to obtain a new user project scoring matrix,which greatly reduced the data sparsity.(2)Due to the collaborative filtering algorithm in the search for the nearest neighbors,need to operate with all users,that leads to low operation efficiency.In order to improve the efficiency of the algorithm,in this paper,considering the user clustering,divide the users of similar likes to the same cluster in the cluster,that can reduce the computational complexity on the maximum extent,improve the scalability of algorithm.(3)The traditional similarity calculation only considers the user rating,which leads to the unreasonable calculation problem.It only calculates the difference between ratings,but does not consider the influence of user attributes on similarity calculation.The user attributes(gender,age,occupation)are introduced into the similarity calculation to optimize the similarity calculation.(4)In order to verify the effectiveness of the proposed optimized algorithm,all experiments in this paper use the Movie Lens100 K dataset,and make experimental comparison with the collaborative filtering algorithm and the filled collaborative filtering algorithm.The results show that the MAE of the algorithm proposed in this paper is improved.