ECOC多分类算法在慕课数据挖掘中的应用  

Application of Multi-Classification Algorithms Based on ECOC in MOOC Data Mining

作　　者：潘丽芳[1] 谢书童[2] 曹秀娟 PAN Lifang;XIE Shutong;CAO Xiujuan(School of Science,Jimei University,Xiamen 361021,China;College of Computer Engineering,Jimei University,Xiamen 361021,China)

机构地区：[1]集美大学理学院,福建厦门361021 [2]集美大学计算机工程学院,福建厦门361021

出　　处：《集美大学学报（自然科学版）》2021年第2期146-151,共6页Journal of Jimei University：Natural Science

基　　金：福建省自然科学基金项目(2018J01538,2020J01707,2020R0066);福建省中青年教师教育科研项目(JAT200266)。

摘　　要：收集并整合多所高校学生的慕课学习行为数据,设计基于数据复杂度的纠错输出编码(ECOC)多分类算法。该算法利用数据复杂度降低多类之间的分类难度,从而提高算法的预测准确度。实验结果表明,在不同高校的慕课数据集的测试中,所设计基于数据复杂度的ECOC分类算法比传统的ECOC算法具有更高的分类准确度和鲁棒性,实现了学生学习成绩多等级的有效预测,为个性化教学奠定了基础。This research collected and integrated the data of learning behaviors of students from many colleges through MOOC platform,and designed ECOC multi-classification algorithms based on data complexity.The algorithms use data complexity to reduce the classification complexity between multiple categories,thereby improving the prediction accuracy of the algorithms.The experimental results showed that the ECOC classification algorithms based on data complexity proposed in this paper have higher classification accuracy and robustness than the classic ECOC algorithms in MOOC data sets of different colleges.The proposed algorithms perform effective prediction of students'academic performance,which lays a foundation to realize the personalized teaching for the students.

关 键 词：ECOC 多分类 慕课 成绩预测 教育数据挖掘 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]