Author:
Cui Pengjia,Yin Boshi,Xu Baichuan
Abstract
AbstractTo improve the accuracy of recommendations, alleviate sparse data problems, and mitigate the homogenization of traditional socialized recommendations, a gated recurrent neural network is studied to construct a relevant user preference model to mine user project preferences. Through the Preference Attention Model Based on Social Relations (PASR), this study extracts user social influence preferences, performs preference fusion, and obtains a Recommendation Algorithm Based on User Preference and Social Influence (UPSI). The study demonstrates that the UPSI algorithm outperforms other methods like the SocialMF algorithm, yielding improved recommendation results, higher HR values, and larger NDCG values. Notably, when the K value equals 25 in Top-K recommendation and using the CiaoDVDs dataset, the NDCG value of the UPSI algorithm is 0.267, which is 0.120 higher than the SocialMF algorithm's score. Considering the user's interaction with the project and their social relationships can enhance the effectiveness of recommendations. Unlike other variants, the UPSI algorithm achieves a maximum hit rate HR value of 0.3713 and NDCG value of 0.2108 in the Douban dataset. In the CiaoDVDs dataset, the maximum hit rate HR value of UPSI is 0.4856, 0.0333 higher than UPS-A, 0.0601 higher than UPS, and 0.0901 higher than UP. Research methods can effectively improve the homogenization problem of traditional socialized recommendations, increase algorithm hit rates and NDCG values. Compared to previous studies, research methods can more fully explore the preference correlation between users, making recommended movies more in line with user requirements.
Publisher
Springer Science and Business Media LLC
Reference29 articles.
1. Williams, A. et al. Quality of internet information to aid patient decision making in locally advanced and recurrent rectal cancer. Surg. J. R. Coll. Surg. Edinburgh Ireland 20(6), 382–391 (2022).
2. Ghai, S. & Trachtenberg, J. Internet information on focal prostate cancer therapy: Help or hindrance?. Nat. Rev. Urol. 6(16), 337–338 (2019).
3. Worthy, J. et al. A critical evaluation of dyslexia information on the internet. J. Literacy Res. 53(1), 5–28 (2021).
4. Drif, A. & Cherifi, H. Migan: Mutual-interaction graph attention network for collaborative filtering. Entropy 24(8), 1084 (2022).
5. Yin, N. A big data analysis method based on modified collaborative filtering recommendation algorithms. Open Phys. 17(1), 966–974 (2019).
Cited by
1 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献