侯璐璐

摘  要： 传统资源协同过滤推荐算法MAE值偏高，因此提出基于Spark架构的艺术学慕课资源协同过滤推荐算法。采用分级响应形式，建立用户?资源评分关系模型，用户对资源的评分减掉该用户评分平均值，完成资源协同过滤相似度计算的优化，引入集成度高的Spark架构，预测用户对资源的评分并生成推荐列表，实现艺术学慕课资源的精准推荐。经过与两种传统算法的对比实验结果可知，研究的算法在不同比例训练集和测试集的情况下，MAE值均低于两种传统方法，说明基于Spark架构的艺术学慕课资源协同过滤推荐算法推荐精度更高，性能更好。

关键词： 协同过滤推荐算法; Spark架构; 艺术学慕课资源; 用户评分预测; 用户?资源评分关系模型; 相似度计算

中图分类号： TN911.1?34; TP319                   文献标识码： A                    文章编号： 1004?373X（2020）03?0162?03

Research on art MOOC resource collaborative filtering recommendation algorithm

based on Spark architecture

HOU Lulu

（Baoji University of Arts and Sciences， Baoji 721013， China）

Abstract： The MAE （mean absolute error） value of traditional resource collaborative filtering recommendation algorithm is slightly higher， so an art MOOC （massive open online course） resource collaborative filtering recommendation algorithm based on Spark architecture is proposed. The user?resource scoring relation model is established in the form of hierarchical response. The average value of the user′s scoring is taken from the user′s scoring for resources， which is then used to optimize the similarity calculation for resource collaborative filtering. The Spark architecture with high integration level is introduced to predict the user′s scoring for resources and generate the recommendation list， thus realizing the accurate recommendation of art MOOC. The results of comparative experiments show that， in comparison with the two traditional algorithms， the MAE value of the proposed algorithm is lower than those of the two traditional methods in different proportion of training sets and test sets， which shows that the art MOOC resource collaborative filtering recommendation algorithm based on Spark architecture has higher recommendation accuracy and better performance.

Keywords： collaborative filtering recommendation algorithm; Spark architecture; art MOOC resource; user scoring prediction; user?resource scoring relation model; similarity calculation

0  引  言

慕课是当今时代下互联网与教育相结合的产物，它实际上是一种大规模开放的在线课程（Massive Open Online Course），是互联网时代下涌现出的一种在线课程的开发模式。传统课程只有几十个或几百个学生，但是一门慕课最多可以容纳十多万人。因此，在互联网中，利用协同过滤进行艺术学慕课资源的推薦[1?2]。协同过滤简单来说，就是利用共同兴趣或者是拥有共同经验人群的喜好大数据资料，来给用户推荐有可能感兴趣的信息，个人通过合作机制给予信息一定程度的回应，利用评分等方法将特别感兴趣的以及特别不感兴趣的资源进行区分，并利用大量的评分记录对信息进行过滤，帮助别人进行信息的筛选。……