IEEE Transactions on Consumer Electronics · 2025 · 60 citations · 39 references
ComputingMarketingConsumer BehaviorE-commerce SystemsRecommender SystemsEvaluation Of Recommender SystemsHybrid RecommendationOnline RecommendationPersonalizationRecommendation EvaluationRecommendation SystemsCommon InformationDiscounted Cumulative GainFederated RecommendationHybrid Recommendation SystemsNormalized Discounted Cumulative GainPersonalized RecommendationPosition EncodingTrustworthy Recommendation
Electronic shopping’s convenience and efficiency make it essential in modern life. In trustworthy personalized consumer recommender scenarios, diverse consumer interests lead to various interactions. Existing methods struggle to capture complex behavior dependencies and shared information across behaviors by exploring multi-behavior interaction sequences. To address this, we propose the Personalized Consumer Federated Recommender System Using Fine-grained Transformation and Hybrid Information Sharing (PCFedRec). Our contributions are as follows: Firstly, we employ the Fine-grained Transformation Module to capture fine-grained heterogeneous dependencies of consumer behaviors and model the behavior semantics of interest. Secondly, we use the Weight Correction Mechanism to reduce suboptimal position encoding and noisy input. Finally, we design the Hybrid Information Sharing Module to learn general consumer interests by leveraging common information across different behaviors. The Normalized Discounted Cumulative Gain is used to evaluate the performance of the recommender systems by considering both the relevance and rank order of the top 10 recommended items. The PCFedRec improves performance by 6.20%, 8.08%, 9.16%, and 7.14% on this metric.
39
Multi-Interest Network with Dynamic Routing for Recommendation at Tmall
Chao Li, Zhiyuan Liu, Mengmeng Wu et al. · 2019 · 354 citations
Interest-aware Message-Passing GCN for Recommendation
Fan Liu, Zhiyong Cheng, Zan Gao et al. · 2021 · 290 citations · Full text