Identifying foreign language learning burnout: latent profiles, cutoff points, and an explainable web-based calculator.

Background: Foreign language learning burnout (FLLB) is prevalent among English as a foreign language learners, which adversely affects students' academic performance and mental health. However, the clear cutoff point and individualized risk prediction tools for FLLB remain lacking, limiting its application in practical identification and assessment. Methods: This cross-sectional study recruited 1,343 Chinese secondary school students and collected data on FLLB, academic stress, engagement, teacher affective support, foreign language enjoyment, English achievement, and demographic characteristics. Latent profile analysis (LPA) was used to identify FLLB risk profiles. The cutoff point was determined using receiver operating characteristic curve (ROC) analysis. Six machine learning models were compared, and their generalizability was evaluated in an external validation set. The best-performing model was selected, interpreted with Shapley additive explanation (SHAP) analysis, and deployed as a web-based calculator. Results: LPA indicated that 14.2% of students belonged to the FLLB high-risk group. The optimal cutoff point was determined as: Exhaustion ≥14, Cynicism ≥3, and Reduced Efficacy ≥3. Logistic regression performed best, with AUC values of 0.903 (internal test set) and 0.813 (external validation set). SHAP analysis revealed that academic stress and foreign language enjoyment were key predictors. Conclusion: This study determined an operational cutoff point for FLLB. A well-performing risk prediction model was then developed and validated, which we subsequently deployed as a web-based calculator. The tool reports FLLB risk probabilities and visualizes the direction and relative contribution of key predictors, thereby providing a reference for efficient FLLB risk screening and subsequent targeted learning support and psycho-educational services.