Research on Computer Intelligent Collaborative Filtering Algorithm for Personalized Network Data Recommendation System

Abstract

In order to meet the protection needs of user privacy data in social networks, this paper proposes a computer intelligent collaborative filtering algorithm for personalized network data recommendation systems. This algorithm predicts user preferences for specific items by utilizing user evaluation information on groups of similar feature items, thereby achieving personalized recommendations. The experimental results show that as the number of project feature selections N increases, the MAE, RMSE, and NDCG5 of the algorithm gradually improve. This is mainly attributed to increasing the number of features under a fixed similarity threshold, which makes the data granularity finer and helps to describe project features more accurately. In the case of a fixed number of project feature selections N, the impact of the number of nearest neighbors s in similar groups on algorithm performance was further studied. The results showed that with the increase of s, MAE, RMSE, and NDCG5 showed a decreasing trend. Although the algorithm suffers from certain losses in recommendation accuracy, it is still within an acceptable range. It is worth noting that due to the system only using generalized data as input, user privacy data is effectively protected. Based on the comprehensive experimental results, this algorithm has significant value in practical applications.