Applying Machine Learning and SHAP Method to Identify Key Influences on Middle-School Students' Mathematics Literacy...

Area of Science:

• Educational Psychology
• Data Science in Education
• Comparative Education

Background:

• PISA 2022 revealed a significant global decline in mathematical literacy among OECD countries.
• East Asian systems (Hong Kong, Macao, Taipei, Singapore, Japan, Korea) showed strong performance, with some improving.
• Understanding factors influencing secondary students' math skills is crucial amid declining trends.

Purpose of the Study:

• To identify key factors affecting middle-school students' mathematical literacy using machine learning.
• To analyze 34,968 samples and 151 features from six East Asian education systems in the PISA 2022 dataset.
• To pinpoint crucial predictors and their influence on mathematical literacy.

Main Methods:

• Utilized machine learning, specifically the XGBoost model, for its high prediction accuracy.
• Applied the SHapley Additive exPlanations (SHAP) method to interpret model predictions.
• Analyzed a large dataset comprising 34,968 samples and 151 features.

Main Results:

• The XGBoost model achieved the highest accuracy in predicting mathematical literacy.
• Fifteen significant predictors were identified, notably mathematics self-efficacy (MATHEFF) and expected occupational status (BSMJ).
• Mathematics self-efficacy emerged as the most influential factor, with varying impacts across individual students.

Conclusions:

• Mathematics self-efficacy is a primary determinant of mathematical literacy.
• Influencing factors and their effects are individualized, requiring tailored educational strategies.
• Four recommendations are proposed to enhance secondary students' mathematical literacy performance.