ZJ-EduFormer: Predicting Supply Chain Student Stress Using Transformer
Jie Zhang, Mcxin Tee, ChenLong Lin, Shen Huili · 2024
In this paper, we propose ZJ-EduFormer, an innovative deep learning model designed to predict stress levels among supply chain management (SCM) students who face unique challenges from multidisciplinary coursework and practical requirements. Our model introduces the Transformer architecture into this domain, featuring a feature projection layer, position encoding mechanism, and dual-layer encoder structure. The multi-head attention mechanism effectively models complex interactions among lifestyle features, while the progressive dimensionality reduction strategy balances performance and generalization ability. Experiments conducted on a dataset of 2000 SCM students demonstrate that ZJ-EduFormer significantly outperforms traditional machine learning models and other deep learning approaches, achieving a 26.1% reduction in RMSE and an R2value of 0.856 compared to the best baseline model. This study expands the application of Transformer in educational stress prediction and provides valuable insights for improving mental health services in SCM education.