摘要
传统CART运算对连续属性处理效率比较低,为此花费较长的运算时间,此外传统CART运算对小量样本数据建立模型不稳定,预测精度不高.为了使用CART运算能高效准确预测M-learning过程中对知识点的掌握程度,本文对CART运算进行改进,首先利用Fayyad边界提高对连续属性最优阀值所用的计算效率,减少分割点个数,减少运算时间.其次对小量样本数据进行基于GB算法的CART建模,多次迭代弱预测器成为较强的预测器,使得小样本数据模型建立更加稳定,提高预测精度.最后实验表明,改进的CART算法对M-learning过程中对知识点的掌握程度的预测,对连续属性处理的速度更快,预测正确率更高,能够提供给学生和教师强有力的决策支持.
The traditional CART operation on the continuous attribute processing efficiency is lower, spending longer operation time, and its model for small sample data is not stable, and prediction accuracy is not high. In order to use the CART algorithm to efficiently predict the mastery of knowledge points in M-learning, this paper presents an improved CART algorithm as the first use of Fayyad border to improve the optimal threshold of continuous attributes used in the calculation efficiency, decrease the number of split points to reduce computing time. Next to the small sample data to CART modeling based on .CB algorithm, multiple iterations weak predictor become stronger predictor, make small sample data model more stable, improve the prediction precision. Finally the experimental results show that the improved CART algorithm of knowledge points has a faster continuous attribute processing, a higher accuracy and can provide powerful decision support for students and teachers.
作者
唐立
李六杏
TANG Li;LI Liu-xing(Department of Information Engineering,Anhui Institute of Economic Management,Hefei 230031,Anhui,China)
出处
《韶关学院学报》
2018年第9期26-31,共6页
Journal of Shaoguan University
基金
安徽经济管理学院院级课题(YJKT1314Q05)