| Summary: | Abstract The interaction and sharing of data based on network users make network information overexpanded, and “information overload” has become a difficult problem for everyone. The information filtering technology based on recommendation could dig out the needs and hobbies of users from the historical behavior, historical data, and social network and filter out useful resource for users in accordance with the needs and hobbies from the accumulation of information resource. Collaborative filtering is one of the core technologies in the recommendation system and is also the most widely used and most effective recommendation algorithm. In this paper, we study the accuracy and the data sparsity problems of recommendation algorithm. On the basis of the conventional algorithm, we combine the user score probability and take the commodity type into consideration when calculating similarity. The algorithm based on user score probability and project type (UPCF) is proposed, and the experimental data set from the recommendation system is used to validate and analyze data. The experimental results show that the UPCF algorithm alleviates the sparsity of data to a certain extent and has better performance than the conventional algorithms.
|
|---|
