Identifying the At-Risk Students Based on Recurrent Neural Network–Informed Logistic Regression Model

• In Event: Measure of Perceptions and Learning Within Instructional Technology Contexts

Abstract

A learning status prediction method was proposed by incorporating the memorizing ability in Recurrent Neural Network into Logistic Regression approach. The model predicts the student latent learning status in an iteratively way by incorporating the previous learning status based on the demographic information and learning activities. The performances of the proposed model were evaluated in terms of predicting the student final performances and latent learning status based on traditional evaluation metrics and the Top-K-Precision/Recall metrics, respectively. Then, the prediction outcomes were interpreted to help the educators to design more appropriate interventions to the at-risk students. The evaluation results show that the proposed model outperforms the baseline models on both predicting the student final performances and identifying the at-risk students.

Authors

• bo pei, University of Florida

• Wanli Xing, University of Florida