A Student Performance Prediction Model Based on Feature Factor Transfer

Temporal information plays an important role in student performance prediction, so two interpretable feature factors, namely enthusiasm and stability, are extracted from online and offline blended temporal learning data, which can characterize students’ learning attitudes. A new transfer predicting model called LGES is proposed. The contrast experiments with and without the feature factors, and the experiments with and without transfer learning are carried out. The results show that the proposed model significantly improves the accuracy and the discrimination ability. Furthermore, it confirms that both transfer learning and the two feature factors can effectively improve prediction performance. The ablation experiment of datasets is conducted. It proves that the model trained on the blended data achieved the best performance, and points that the importance of offline data in identifying truly failing students.