基于多分类器的迁移Bagging习题推荐

文章编号:10019081(2013)07195005

doi:10.11772/j.issn.10019081.2013.07.1950

摘 要:

针对协同过滤(CF)推荐方法用户的历史信息不足等问题，提出基于多分类器的迁移Bagging习题推荐算法。主要思路是把推荐问题投入迁移学习框架，将待推荐习题的用户作为目标域，从中搜索相似历史信息的用户作为辅助域，帮助训练目标域以得到更准确的分类结果。实验结果表明，所提方法在习题推荐库及公开数据上，比协同过滤算法性能提高了10%~20%；比单分类器Bagging迁移算法性能提升了5%~10%。该方法在一定程度上解决了习题推荐系统中存在的冷启动和数据稀疏问题，也可推广到商品推荐等电子商务平台。

关键词： 迁移学习；Bagging；协同过滤;推荐系统；计算机辅助教学

中图分类号：TP391.7文献标志码:A

英文标题

Online transferBagging question recommendation based on hybrid classifiers

英文作者名

WU Yunfeng, FENG Jun*, SUN Xia, LI Zhan, FENG Hongwei, HE Xiaowei

英文地址(

School of Information Science and Technology, Northwest University, Xian Shaanxi 710069, China英文摘要)

Abstract:

Traditional Collaborative Filter (CF) often suffers from the shortage of historic information. A transferBagging algorithm based on hybrid classifiers was proposed for question recommendation. The main idea was that the recommendation and prediction problem were cast into the framework of transfer learning, then the users demand for recommend questions were treated as target domain, while similar users who had applicable historic information were employed as auxiliary domain to help training target classifiers. The experimental results on both question recommendation platform and popular open datasets show that the accuracy of the proposed algorithm is 10%-20% higher than CF, and 5%-10% higher than single Bagging algorithm. The method solves cold startup and sparse data problem in question recommendation field, and can be generalized into production recommendation on Ecommerce platform.

Traditional Collaborative Filtering (CF) often suffers from the shortage of historic information. A transferbagging algorithm based on hybrid classifiers was proposed for question recommendation. The main idea was that the recommendation and prediction problem were cast into the framework of transfer learning, then the users demand for recommend questions were treated as target domain, while similar users who had applicable historic information were employed as auxiliary domain to help training target classifiers. The experimental results on both question recommendation platform and popular open datasets show that the proposed algorithm improves 10%~20% accurate rates compared with CF, and increases 5%~10% than that of single bagging algorithm. The method solves cold startup and sparse data problem in question recommendation field, and can be generalized into production recommendation in Ecommerce platform.

英文关键词Key words:

transfer learning; Bagging; Collaborative Filtering (CF); recommendation system; Computer Assisted Instruction (CAI)



0 引言

随着Internet的发展,网络辅助教学系统(Online Aided Teaching System, OATS)为传统教育提供了随时随地的帮助。OATS通常包括在线项目实践、在线习题测试、在线问答等功能。例如，通过在线教学平台，Sturm大学法学院的各地考生能随时参加该院组织的一场专业考试［1］。哈尔滨工业大学的C语言教学系统能完成在线考试、自动评分和统计工作［2］。但是，目前大部分的OATS提供的只是简单辅助教学或者学资源管理的功能，即给每个学生提供相同的习题和资源。然而不同学生的个性、背景不尽相同，对不同习题和知识点的理解力及偏爱也不一样。为了达到个性化教学、提高学习效率的目的，对不同学生推荐有针对性的习题和个性化的教学资源成为OATS系统的最新目标。