Interpretable AI Model to Determine the Key and Difficult Points of Test Papers

Abstract

Test paper analysis plays an important role in the field of educational statistics and evaluation. The rapid development of machine learning has brought new opportunities for test paper analysis. This study uses decision tree and XGBoost method, combined with SHAP interpretation framework, to analyze the key and difficult questions in the university physics test paper, and quantify how each key and difficult question affects students' passing rate and excellence rate. The research results show that the performance of XGBoost method is better compared with decision tree method. Using explainable artificial intelligence technology to show how each question affects students' test performance can help the teachers explain the pass rate and excellent rate, and provide a reference for the application of explainable artificial intelligence in test paper analysis and evaluation.