计算机应用研究2025，Vol.42Issue(9)：2660-2668,9.DOI:10.19734/j.issn.1001-3695.2025.03.0045
基于图采样的多样性需求感知推荐模型
Graph sampling based recommendation model for diversity needs awareness
摘要
Abstract
Existing diversified recommendation methods ignore individual diversity needs differences,causing insufficient adap-tation of recommendation diversity.To address this issue,this study proposed a graph sampling based recommendation model for diversity needs awareness,which achieved personalized recommendation diversity by perceiving diversity needs intensity and optimizing graph structures.It first quantified users' diversity needs intensity through time-decay weighted item dissimilarity.Secondly,the method developed an adaptive graph sampling strategy based on needs intensity,determining sampling frequency according to users' diversity needs levels.It applied a greedy algorithm to iteratively select nodes with the highest neighborhood dissimilarity,constructing a diversity-needs-aware subgraph for graph learning.Then,the model used a graph neural network to learn the diverse interest representations of users from the diversity-needs-aware subgraph.Finally,the model predicted the in-teraction probability between users and items through dot product operations and generated diversified recommendations.The experimental part used two public datasets for verification.The proposed model improves accuracy metrics by approximately 3％and diversity metrics by 5％.These results demonstrate that incorporating users' diversity needs effectively achieves a better accuracy-diversity trade-off.关键词
图采样/多样化推荐/多样性需求/图神经网络Key words
graph sampling/diversified recommendation/diversity needs/graph neural network分类
信息技术与安全科学引用本文复制引用
徐建民,鲁平,张雄涛..基于图采样的多样性需求感知推荐模型[J].计算机应用研究,2025,42(9):2660-2668,9.基金项目
国家社会科学基金资助项目(23BTQ092) （23BTQ092）
