基于LDA主题模型和学习者行为特征的协同过滤个性化学习资源推荐方法  

Method of Collaborative Filtering Recommendation for Learning Resources based on LDA and Learners’ Behavior

作　　者：董露露 马宁[1] DONG Lu-lu;MA Ning(Anhui Open University,Anhui Hefei 230022,China)

机构地区：[1]安徽广播电视大学

出　　处：《贵阳学院学报（自然科学版）》2019年第4期71-78,共8页Journal of Guiyang University：Natural Sciences

基　　金：国家开放大学一般科研项目:基于主题模型的远程教育个性化学习资源推荐应用研究(项目编号:G18F1805Y);安徽高校自然科学重点项目:远程教育中基于动态兴趣模型的个性化学习资源推荐研究(项目编号:KJ2019A0968)

摘　　要：随着在线学习方式的普及,网络学习资源急剧增加,亟需为学习者提供高质量的个性化资源推荐服务。协同过滤推荐是最常用的推荐方法之一,但传统协同过滤存在数据稀疏性问题导致其难以得到大规模应用。为此,提出一种基于主题模型和学习者行为特征的协同过滤资源推荐方法。首先,利用LDA主题模型构建资源和学习者的主题向量;其次,采用线性加权的方式计算学习者相似度以寻找近邻;然后,分析学习者行为特征,完成动态兴趣建模;最后,利用近邻预测学习者感兴趣的资源。实验结果表明,所提方法能有效改善个性化学习资源推荐效果。Learning resources on the internet increase greatly with the popularity of e-learning and the highly recommendation service is needed urgently for the learners.Collaborative filtering is one of the most widely used methods,but it has the problem of data sparseness,which makes it difficult to achieve large-scale applicaiton.Based on this,a collaborative filtering approach based on LDA and learners’behavior is proposed in this paper.Firstly,construct the topic vectors of resources and learners by LDA.Secondly,calculate the learners similarity to search for the nearest neghibours by linear weighted method.Then,analyze the learners’behaviors to describe learners’interests more completely.Finally,predict learners’interested resources by nearest neighbor.Experiment results show that the proposed approach can effectively improve the personalized resources recommendation effect.

关 键 词：LDA 学习者行为 协同过滤 个性化 资源推荐 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]