关键词: 特征网络, 群集特征, 稀疏化, 凸非负矩阵分解, PSR 评价体系, 分级模型

Abstract: Aiming at the lack of fusion analysis on the mixed data of online teaching and traditional classroom， the improved sparse convex non-negative matrix factorization algorithm was used to effectively extracting the feature clusters of student learning behavior data. Based on the weights of cluster features，multi-level feature indicators were selected in turn to construct a PSR evaluation index system on three levels of evaluation layer，cluster layer and feature layer. According to the evaluation index system，the quality value of individuals was calculated based on the comprehensive weighting method，and students were graded. The grading results were in accordance with the normal distribution and basically consistent with the grading distribution of students' final examination scores. Thus the validity of the model and method proposed in this paper is proved.

Key words: feature network, cluster characteristics, sparsification, convex non-negative matrix factorization, PSR evaluation system, hierarchical model