协作学习环境中的协同过滤算法

摘要

用户数与学习资源数的增加导致在线协作学习不能更好地体现其个性化，协同过滤推荐算法推荐精度下降。为解决该问题，提出一种适合在协作学习环境下应用的协同过滤推荐算法。将协作学习环境中学习者与学习资源属性作为学习推荐中相似度计算的一部分，将属性相似度与学习资源评分相似度作为综合相似度，加入时间函数保证最近发生的兴趣有更大的权重，最终完成推荐。为体现客观性，实验结果分别选取在不同学习者数、不同学习资源数以及不同稀疏程度的资源评分矩阵下的平均绝对误差(MAE)。实验结果表明，与基于项目属性的用户聚类协同过滤推荐算法相比，在各种不同的实验情况下，该算法的实验结果 MAE更小，推荐精度更高。%With the increasing number of users and resources,online collaborative learning cannot better reflect the personaliza-tion,and the recommendation accuracy of collaborative filtering recommendation algorithm is declining.To deal with these prob-lems,a collaborative filtering recommendation algorithm suitable for collaborative learning environment was proposed.The pro-perties of learners and resources in the collaborative learning environment were made as part of similarity calculation,and recom-mendation was achieved by comprehensive similarity and function of time.The comprehensive similarity consisted of attribute similarity and rating similarity.The function of time assured that a larger weight was occupied by the recent interest.To reflect objectivity,the experimental results employed mean absolute error (MAE)in different number of learners,different number of resources,and rating matrix on different sparsities.The simulation results indicate that the proposed algorithm has less MAE and more accurate recommendation effectiveness for the system in different conditions of the experiments,in comparison with collaborative filtering recommendation algorithm based on user clustering of item attributes.