基于改进谱聚类算法的学生学业预测模型

Prediction of student academic performance via unproved spectral clustering algorithm

预测学生学业成绩是教育数据挖掘(EDM)领域最重要的研究课题之一。高职院校计算机相关专业学生在诸如程序设计和数据结构等课程上面临各种各样的困难,这类课程的成绩不及格是大部分学生重修甚至辍学的主要因素。因此,使用EDM分析学生在校学习行为数据,达到提前预测学生学业成绩的目的,实现及时学业预警,并帮助学生在未来取得更好的成绩。针对谱聚类在处理具有复杂统计特性数据集时的不足,设计了一种通过Siamese网络从数据中学习距离度量构造相似性矩阵的改进谱聚类算法,再把学生成绩相关数据进行聚类,获取学生学业表现预测模型。实验结果表明,模型预测学生成绩的准确率为98.3%,高于k-means和传统谱聚类模型。

Predicting students’academic performance is one of the most important research topics in the field of Educational Data Mining(EDM).Computer-related majors in higher vocational colleges are faced with various difficulties in such courses as program design and data structure,and the failure rate of such courses is the main factor that leads most students to retake or even drop out of school.Therefore,EDM is used to analyze student behavior data collected from educational environments to predict students’academic performance,to timely academic warning,and help them achieve better results in future courses.Aiming at the shortcomings of spectral clustering in dealing with datasets with complex statistical characteristics,this paper designs an improved spectral clustering algorithm that constructs a similarity matrix by learning distance metrics from data through the Siamese network and then clusters the data related to students’academic performance to obtain the prediction model of students’academic performance.The experimental results show that the accuracy of the proposed model in predicting students’grades is 98.3%,which is higher than k-means and traditional spectral clustering model.