Research on diversity and accuracy of the recommendation system based on multi-objective optimization

As the information industry and the Internet develop rapidly, the use of big data enters people's vision and attracts attention. It makes the recommendation system come into being how to quickly extract the desired information from the excessive information. In the recommendation system, user-based collaborative filtering algorithm has become a research hotspot. Existing researches focus on improving collaborative filtering recommendation algorithm by using the kernel method, but still face the cold start problem, the diversity problem, the data sparsity problem, the concept drift problem and more others. To solve these problems, this paper proposes the user-based collaborative filtering based on kernel method and multi-objective optimization (MO-KUCF) which introduces kernel density estimation and multi-objective optimization. It can be increasing diversity of the recommendation systems, improving concept drift in dynamic data and the accuracy and diversity of the recommendation system. The dataset used in this article is the Netflix dataset. It analyzes the MO-KUCF algorithm with the user-based collaborative filtering (UCF) and user-based collaborative filtering based on kernel method (KUCF) by the mean absolute error (MAE). The MAE is compared with the internal user diversity I-u index, and the pre-processed data set is divided into the training set and the test set, which are provided to the recommendation system for recommendation and evaluation. The results show that the accuracy of MO-KUCF improves by 5.6%, and the diversity also increases with decreasing values. Combining multi-objective optimization techniques with kernel density estimation methods can improve the diversity of recommendation systems effectively and solve the concept drift problem to achieve the purpose of improving system accuracy.

Automated summary

With the rapid development of the information industry and the Internet, the use of big data has attracted attention, leading to the emergence of recommendation systems that aim to quickly extract desired information from vast amounts of data. User-based collaborative filtering algorithm has become a research focus in this field. However, existing research mainly focuses on improving collaborative filtering recommendation algorithms using kernel methods, but still face various challenges such as cold start, diversity, data sparsity, and concept drift. To address these challenges, this paper proposes a user-based collaborative filtering algorithm based on kernel methods and multi-objective optimization (MO-KUCF). By introducing kernel density estimation and multi-objective optimization, the proposed algorithm enhances recommendation system diversity, helps deal with concept drift in dynamic data, and improves the accuracy of recommendations. The Netflix dataset is used for empirical analysis, comparing the MO-KUCF algorithm with user-based collaborative filtering (UCF) and user-based collaborative filtering based on kernel methods (KUCF) using mean absolute error (MAE). The results show that MO-KUCF improves accuracy by 5.6% and increases diversity. By combining multi-objective optimization techniques with kernel density estimation methods, the proposed algorithm effectively improves recommendation system diversity and solves the concept drift problem, thereby enhancing system accuracy.