基于注意力的深度协同在线学习资源推荐模型

An Attention-based Deep Collaborative Filtering Model for Online Course Recommendation

推荐可以看作是一个匹配问题,即为适当的用户匹配适当的项。针对学习平台和课程资源数量剧增以及在线资源分散使得课程推荐质量不佳等问题,将注意力机制和深度学习融入课程推荐问题中,提出一个基于注意力的深度协同在线学习资源推荐模型来为高阶课程集关系进行建模。该模型结合学习者信息和课程资源特征,学习用户和课程的隐性线性特征和非线性特征,进行多模态特征拼合,融入注意力机制思想区分不同成对项目集对预测结果的贡献程度,以提升模型表示用户和课程的准确性,提高推荐性能。通过爬取慕课网(MOOC)上2014-2019年的学习数据进行实验,结果表明提出的模型在数据集userlabel08rl上多项评价指标要明显优于其它推荐算法。

A recommendation can be thought as a matching problem,i.e.,matching the appropriate item for the appropri-ate user.Aiming at the problems such as the increasing number of learning platforms and course resources and the poor quality of course recommendation caused by online resource dispersion,this paper integrates the attention mechanism and deep learning into the course recommendation,and proposes an attention-based deep collaborative filtering model for online learning resource recommendation so as to model the high-level course set relation.This model combines the characteristics of learner information and course resources to learn the invisible linear features and nonlinear features of the user and the course,and performs multi-modal feature matching,in order to improve the accuracy of the model in representing the user and the course and the recommendation performance.Through the experiment of crawled learning data from 2014 to 2019 on the MOOC network(MOOC),the results show that the proposed model in this paper is significantly better than other recommendation algorithms in multiple evaluation indexes on the real data set userlabel08rl.