The College English Test Band 4 (CET-4) is a significant examination in Chinese higher education, used to assess students’ English proficiency. This study focuses on comparing prediction models for evaluating CET-4 scores, with a specific emphasis on applying machine learning (ML) techniques. This research introduces a hybrid Intelligent Crow Search Optimized Adaptive Decision Tree (ICSO-ADT) model to identify the most effective approach for predicting CET-4 performance. The model leverages variables such as prior academic performance, study habits, and psychological factors to train and validate predictions using a sample of university students’ CET-4 scores. Data preprocessing is carried out using normalization techniques, ensuring consistency and quality. Evaluation metrics, including accuracy and error rates, are employed to assess the model’s predictive capabilities. The results demonstrate that the ICSO-ADT approach outperforms existing techniques, achieving a prediction accuracy of 95% with the lowest error rates. This highlights its superior ability to capture non-linear relationships between input variables and CET-4 outcomes. The findings are significant for educators and administrators, offering a data-driven method for enhancing English teaching strategies and identifying students in need of additional support. Future work will explore more complex predictors and further optimize model parameters to achieve even greater prediction accuracy and efficiency, advancing the field of English language education and assessment.
Jingjing Xiong (Tue,) studied this question.