基于机器学习的学生成绩预测及教学启示  

Students' Performances Prediction and Teaching Cogitation Based on Machine Learning

作　　者：吕品[1] 于文兵[1] 汪鑫[1] 计春雷[1] LYU Pin;YU Wen-bing;WANG Xin;JI Chun-lei(Shanghai Dianji University,Shanghai 201306,China)

机构地区：[1]上海电机学院,上海201306

出　　处：《计算机技术与发展》2019年第4期200-203,共4页Computer Technology and Development

基　　金：上海市教育科学研究项目(C17014/17AR04);上海电机学院重点教研教改项目(A1-0224-17-009-05);上海电机学院计算机科学与技术优势学科(16YSXK04)

摘　　要：利用学习分析技术挖掘在线学习特征是理解与优化教学过程、实现教学决策和学业预警的重要依据。在采集在线学习者的人口统计信息、学习背景、家长参与以及学习者的行为特征等信息的基础上,首先使用感知机、支持向量机和神经网络等分类算法,分别构造了不同的学习成绩预测模型;通过比较模型的准确度、召回率、F值,误分类样本数量和精确度,最终选择基于支持向量机的成绩预测模型。其次,通过分析模型参数,得出了影响学习成绩的主要因素是学习者参与小组讨论、课堂举手、访问与课程相关资源以及浏览通告等学习者行为特征的结论。最后,提出教师应该关注学生的学习行为特征,合理运用价值动机理论和内隐智力信念调节机制,激发学生的学习投入和学习动力的教学启示。The use of learning analytic technologies to mine online learning features is an important basis for understanding and optimizing the teaching process,realizing teaching decisions and learning early warning.On the basis of collecting online learners’ demographic information,learning background,parents ’ participation and learners’ behavior characteristics,we firstly construct different learning performance prediction models by perceptron,support vector machine and neural network. By comparing the precision,recall,F-score,the number of misclassified samples and accuracy of these models,the support vector machine is selected as the final performance prediction model.Secondly,it is concluded that learners’ participation in group discussions,raising hands in class,access to course- related resources and browsing notices are the main factors affecting their academic performance.Finally,it is suggested that teachers should pay close attention to the characteristics of students’ learning behavior,rationally apply the theory of value motivation and the mechanism of implicit intelligence belief to stimulate students’ learning engagement and impetus.

关 键 词：教育数据挖掘 感知机 支持向量机 神经网络 学习成绩预测 教学启示 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]