垂直学习社区基于学习兴趣与风格的社会化推荐算法

Research on Social Recommendation Methods Based on Learning Interest and Style in Vertical Learning Community

垂直学习社区包含了海量的学习资源,出现了信息过载现象,个性化推荐是解决这个难题的方法之一.但垂直学习社区中评分数据稀疏而文本、社交信息丰富,传统的协同过滤推荐算法不完全适用.基于用户产生的文本和行为信息,利用作者主题模型构建新的用户学习兴趣相似度衡量模型;根据用户交互行为信息综合考虑信任与不信任因素构建用户全面信任关系计算全面信任度;通过分析用户多维度学习行为模式,自动识别用户学习风格;最后提出融合兴趣相似度、全面信任度及学习风格的社会化推荐算法.用垂直学习社区网站CSDN实际数据集进行了实验分析.结果表明本文提出的推荐方法能更好向用户推荐其感兴趣的学习资源,有效地提高了推荐精度,进而提高用户学习效果.

With the growing number of users of online education,The vertical learning community contains a huge amount of learning resources,which brings certain difficulties to users’choices.There is an information overload phenomenon,and personalized recommendation is one of the ways to solve this problem.However,in the vertical learning community,the scoring data is sparse and the text and social information are rich.The traditional collaborative filtering recommendation algorithm is not fully applicable.Based on the text information and behavior information generated by users,this paper proposes a newuser learning interest similarity measurement model by using author topic model.Then this paper constructs user comprehensive trust relationship based on interaction behavior information.What’s more,identifying user learning style automatically through analyzing user multi-dimensional learning behavior pattern.Finally,we propose a social recommendation method that integrates user learning interest,comprehensive trust relationship and learning style.The experimental data analysis was carried out with the actual data set of CSDN.The results showthat the proposed method can better recommend the learning resources that are of interest to users,effectively improve the recommendation accuracy and improve the user learning effect.