计算机应用研究2025，Vol.42Issue(11)：3284-3290,7.DOI:10.19734/j.issn.1001-3695.2025.03.0095
基于评分预测与图模型扩散的推荐方法
Recommendation method based on rating prediction and graph model diffusion
摘要
Abstract
To address the issues of data sparsity and limited recommendation scope in collaborative filtering algorithms,this paper proposed a recommendation method based on rating prediction and graph model diffusion,named SIRR.Firstly,it designed a dynamic switching mechanism based on the number of user ratings to predict ratings for unrated items,aiming to address the data sparsity problem.Secondly,it improved the accuracy of similarity computation and the robustness of the collaborative filtering algorithm using regularized cosine similarity.Finally,to overcome the limitation of localized recommendations,it applied a weighted random walk on the graph to expand the recommendation scope,enhancing coverage.To balance recommendation accuracy and diversity,it achieved an optimization by integrating rating weights.It validated the effectiveness of regularized cosine similarity on two datasets of different types.The proposed method was compared with three baseline algorithms on three datasets with varying sparsity levels.Simulation re-sults show that SIRR performs well across all evaluation metrics.It provides an effective solution to data sparsity and local recommen-dation problems.关键词
局部推荐/评分预测/正则化余弦相似度/图的加权随机游走/评分权重Key words
local recommendation/score prediction/regularized cosine similarity/weighted random walk of graph/rating weight分类
计算机与自动化引用本文复制引用
王柳,陈学斌,高远,马凯光,赵桐..基于评分预测与图模型扩散的推荐方法[J].计算机应用研究,2025,42(11):3284-3290,7.基金项目
国家自然科学基金资助项目(U20A20179) （U20A20179）
