一种基于隐式信任感知的MOOCs推荐方法

目的/意义]随着MOOCs迅猛发展和普及,如何利用智能推荐技术为学习者从海量的MOOC中"寻找最佳课程"成为MOOC发展中需要解决的重要课题。方法/过程]基于自我知觉理论和学习行为投入框架,充分利用学习行为日志和评分数据挖掘学习者之间的隐式信任关系,并通过信任传播建立MOOC社区信任网络,从而构建动态结合兴趣和隐式信任感知的混合推荐方法。为解决数据稀疏问题,提出基于信任的联合概率矩阵分解模型(TA-PMF),将课程评分矩阵、信任关系矩阵的分解相结合来挖掘用户及课程潜在特征,进而实现评分预测。结果/结论]真实数据集测试结果表明,与显性评分值相比,学习行为投入信息对信任度构建贡献权重达到0.7;TA-PMF方法对MOOC推荐具有较好的适用性,且能在一定程度上缓解冷启动问题。

A Recommendation Method for MOOCs Based on Implicit Trust-Awareness

Purpose/significance]With the rapid development and popularization of MOOCs,it has become an important issue as how to use intelligent recommendation technology to find the best course for learners from massive MOOCs.Method/process]Based on self-perception theory and learning behavioral engagement framework,a hybrid recommendation method combining interest and implicit trust awareness dynamically is constructed through making full use of learning behavior log data and rating data to find out the implicit trust relationship among leaners and establishing MOOC community trust network through trust propagation.In order to solve the problem of sparse trust data,a probabilistic matrix factorization model based on trust-aware(TA-PMF)is proposed,which combines the factorization of learners’trust relationship matrix and the learner course rating matrix to mine the potential features of users and courses.Result/conclusion]Through collecting and testing data from 100 courses from China’s MOOC platform,it is found that more potential trust relationships can be found from learners’behavioral engagement information than explicit scoring,which is as high as 0.7.Moreover,TA-PMF algorithm integrated trust and rating information is quite compatible with MOOC recommendation,which can help alleviate the problems of cold start and new users to a certain extent.