基于Lasso-logistic回归和随机森林模型的院校评价结果影响因素研究

Research on Influencing Factors of College Evaluation Results Based on Lasso-logistic Regression and Random Forest Model

将Lasso-logistic模型引入院校评价结果影响因素的厘定研究,以教育部高职“双高”院校为研究个案,相较于支持向量机、决策树等模型,Lasso-logistic能够更高效地压缩筛选出入选“双高”院校的关键解释变量,而且总体预测准确率近80%,模型外推性良好,通过随机森林模型进一步验证了Lasso-logistic结果的合理性.实证结果表明,专业、教师、学生标志性成果是核心影响因素,建校时间、教师数、生师比等因素未产生实质性影响.

Introducing the Lasso-logistic model into the research on the determination of the factors affecting the evaluation results of institutions,taking the Ministry of Education’s“double-high colleges”as a case study.Compared with support vector machines and decision trees,Lasso-Logistic is more efficient The key explanatory variables selected for the“double colleges and universities”were compressed and selected,and the overall prediction accuracy was nearly 80%.The model has good extrapolation.The random forest model further verified the rationality of the Lasso-logistic results.The empirical results show that major,teacher,and student landmark achievements are the core influencing factors,and factors such as school establishment time,number of teachers,student-teacher ratio and other factors have no substantial impact.