Research And Implementation Of The Cold Start Problem In Recommender Algorithm

The recommendation technique can make up for the shortcomings of information retrieval technique.In context of information overload,it helps filter information and better meet user's personalized needs.Collaborative filtering algorithm has been widely applied to make recommendations.However,with the rapid growth of the number of items and users in the recommender system,the cold-start problem becomes increasingly apparent which seriously affects the quality of recommendation.Therefore,in order to make the recommender algorithm show good performance,the paper improves the traditional one and the improvement is divided into the following four aspects.Firstly,aiming at data sparsity that is likely to affect clustering and recommendation results,the paper prefills the rating matrix based on the similarity of item attributes,so as to effectively alleviate data sparsity.Secondly,to deal with user cold-start,the paper introduces user demography and modifies similarity calculation combined with both demography and ratings.The system can dynamically adjust the proportion of the two according to different situations of different users.When a new user comes to the recommender system for the first time,optimized similarity calculation can help find nearest neighbors of the new user based on demographic similarities and then make recommendations.Next,in order to solve the problem of poor real-time,k-means is applied to cluster users offline.The paper optimizes the selection of initial clustering centers to improve the stability of the algorithm.Then,it uses optimized k-means algorithm to do user clustering offline,which is to divide users with different interests into different categories.When a new user comes to the recommender system,it does online user partitioning.The principle of "offline clustering,online partitioning" can effectively improve the efficiency of finding nearest neighbors and enhance the real-time of the recommender system.Finally,to solve the problem of poor diversity,the paper proposes a novel method to punish the score of items with high popularity by stages,which is to improve diversity as much as possible at the expense of minimal accuracy.The improved recommender algorithm can work normally in the new user cold-start environment.The paper verifies the proposed algorithm using MovieLens.And the experimental results show that the optimized recommender algorithm can effectively solve problems of data sparsity and user cold-start.Compared with other recommender algorithms,the optimized one improves the accuracy of recommendation.