基于混合兴趣主题模型的推荐方法
A recommendation method based on mixed interest topics model
邱云飞 1田丰维1
作者信息
- 1. 辽宁工程技术大学 软件学院,辽宁 葫芦岛 125105
- 折叠
摘要
针对跨领域项目推荐过程中用户兴趣稀疏造成的推荐冷启动问题,提出一种基于混合兴趣主题模型兴趣领域潜在狄利克雷分布(PA-LDA)的推荐方法.PA-LDA 使用兴趣潜在狄利克雷分布(P-LDA)模块挖掘用户历史行为数据,生成关于目标项目中兴趣主题的概率分布,综合考虑主题和项目内容词对兴趣的影响进行参数估计建模,得到用户对目标项目的兴趣评价.PA-LDA使用领域潜在狄利克雷分布(A-LDA)得到领域对项目目标的兴趣评价,混合两类兴趣评价,使用top-k方法推荐目标项目.在EdX和GCSE两组真实数据集上进行实验,验证方法的有效性和准确性.研究结果表明:PA-LDA可以有效解释用户兴趣和领域兴趣对项目推荐的作用原理,实现多维领域推荐的兴趣特征捕捉,提升推荐的适应性与准确性.
Abstract
To solve the cold start problem caused by user interest sparsity in cross-area project recommendation,this paper proposes a recommendation method on mixed interest topic model PA-LDA.PA-LDA uses the P-LDA module,which generates the interest topic distribution to target project by mining users'historical behavior data.Then P-LDA employs conduct parameter estimation to build model by the interaction between the topics and the content words,which helps to measure the users'interest on the target project.PA-LDA uses A-LDA module to measure the area interest on the target project.PA-LDA employs top-k method to recommend the target project based on the result of the two mixed interest measurements.The effectiveness and efficiency of our method are verified by experiments on two real data sets EdX and GCSE.The research can effectively explain the principles of effect on recommendation by user interest and domain interest.It also realizes the interest feature capture in multi-dimensional area recommendation,which improves the adaptability and accuracy of recommendation.
关键词
主题模型/用户兴趣/领域兴趣/兴趣混合/top-k推荐Key words
topic model/user interest/area interest/interest fusion/top-k recommendation引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据