基于过采样融合集成学习的研究生培养质量评估方法研究

A Study on the Cultivation Quality Evaluation Method of Postgraduate Based on Oversampling Fusion Ensemble Learning

针对研究生培养质量评估中存在的主观性强、数据类别不平衡问题,文章提出一种基于过采样与集成学习的研究生培养质量预测算法(K-means SMOTE Random_Stacking,KSRS)。首先,通过K-means SMOTE算法调整数据集样本分布,使其各类别平衡;其次,基于平衡数据集采用Random_Stacking算法构建研究生培养质量预测模型;最后,利用模型对研究生培养质量进行预测。根据真实的研究生培养数据验证所提模型的有效性。结果表明:对比其他分类算法,KSRS算法在指标召回率、F1值、精确率上均有较大提升,为研究生培养质量评估提供一种科学有效的方法参考。

To address the problems of strong subjectivity and unbalanced data categories in graduate training quality assessment,the paper proposes a graduate training quality prediction algorithm(K-means SMOTE Random_Stacking,KSRS)based on oversampling and integrated learning.Firstly,the sample distribution of the data set is adjusted by K-means SMOTE algorithm to make its categories balanced;secondly,a graduate training quality prediction model is constructed based on the balanced data set using Random_Stacking algorithm;finally,the model is used to predict the graduate training quality.The validity of the proposed model is verified based on the real postgraduate training data.The results show that,compared with other classification algorithms,the KSRS algorithm has a greater improvement in the index recall,F1 value and accuracy rate,which provides a scientific and effective method reference for graduate training quality assessment.