Recommendation Algorithm Combining Tag Data and Naive Bayes Classification

Abstract

Collaborative filtering algorithm is the most popular algorithm applied to recommendation systems. However, it has been plagued by the cold start problem which seriously affects the effectiveness of recommendation. Aiming at users' cold start problem, we proposed a recommendation algorithm which combined tag data and Naive Bayes classification. Tag data was used to represent the users' attribute characteristics, and new and old users were matched by Naive Bayes classifier, which utilized the similar user groups to infer the interests of new users. After determining the categories of new users, we calculated the average rating information of users in this category for items to achieve Top-N recommendation. Experiments showed that the algorithm can achieve better RMSE in the problem of the cold start of users, and the recommendation accuracy was significantly improved.