针对远程教育中传统的集成算法通常以批模式方式运行而导致其在连续生成数据的情况下不可用的问题，提出了一种基于分类器组合增量集成的远程教育学生表现预测算法。首先，简要介绍了三种备受关注的集成分类算法：朴素贝叶斯的增量版本、1‐NN和WINNOW算法；然后，在训练数据集上利用三种算法产生各自的假说；最后，将三种假说进行集成，并利用投票方法预测学生的表现。在希腊远程教育大学“信息”课程提供的训练集 HOU 上的实验结果表明，相比其它几种较好的分类器，该文算法取得了更好的分类精度和更少的训练时间，为教师提供了强有力的学生表现预测工具。%Traditional integrated algorithms usually run as batch mode in remote education ,for the issue that it is una‐vailable under the condition with data generated continuously ,a student's performance predicting algorithm based on combi‐nation incremental integration(CII) with classifiers is proposed .Firstly ,the three popular ensemble classifiers incremental version of simple Bayesian ,1‐NN and WINNOW algorithm are introduced .Then ,the three algorithms are used to generate each hypothesis .Finally ,three hypothesizes are integrated and voting method is used to predict student's performance .Ex‐perimental results on training set HOU supplied by information course at Greek University of Distance Education show that proposed algorithm has higher classification accuracy and less training time than several advanced classifiers ,which indicates that it provides a powerful prediction tool of student's performance .
展开▼
