文献类型：期刊文献

英文题名：Fuzzy topic modeling with learnable thresholds for aspect personalized video recommendation

作者：Li, Te[1];Guo, Yi[1];Fu, Jiaojiao[1]

机构：[1]East China Univ Sci & Technol, Sch Informat Sci & Engn, Meilong Rd 130, Shanghai 200237, Peoples R China

年份：2026

卷号：68

期号：1

外文期刊名：KNOWLEDGE AND INFORMATION SYSTEMS

收录：;EI(收录号：20261720598972);WOS:【SCI-EXPANDED(收录号:WOS:001745273800002)】；

语种：英文

外文关键词：LDA; MLP; Recommendation algorithm; Fuzzy logic; LDA; MLP; Recommendation algorithm; Fuzzy logic

摘要：Aspect recommendation enables the precise learning of user preferences from multiple angles. However, conventional approaches that rely on fixed thresholds for recommendation decisions fail to meet users' personalized needs. Additionally, traditional discrete rating mechanisms are insufficient to capture nuanced user interests. We introduce a Dynamic Threshold-Enhanced Fuzzy Double-Layer Latent Dirichlet Allocation (DynaFuzz-DLDA) model to address these challenges. This unsupervised framework integrates fuzzy logic and adaptive thresholds into a double-layer topic modeling architecture. Firstly, the DynaFuzz-DLDA model utilizes the LDA model and aspect-term association graph-based corpus to extract topic information from videos based on user comments. Secondly, this study proposes an improved association rule method to address the data imbalance problem in the dataset, thereby preventing the model from overfitting. Thirdly, user-video rating information is probabilistically modeled, and a Fuzzy Latent Dirichlet Allocation (FLDA) topic model is constructed based on the LDA model to extract finer-grained potential user interest features. Finally, this paper proposes an adaptive threshold adjustment strategy based on the Multilayer Perceptron (MLP) neural network to enhance the performance of recommendations. Experiments are conducted on real-world datasets, and the results show that the DynaFuzz-DLDA model demonstrates competitive performance against these baselines in terms of overall model performance.