Main content
Modeling Student Math Achievement Across Countries with Machine Learning Using TIMSS 2019
Date created: | Last Updated:
: DOI | ARK
Creating DOI. Please wait...
Category: Project
Description: Children’s early math skills are critical for future academic success. To profile the most important predictors of student math achievement, this study applies empirically driven supervised machine learning (ML) techniques to the Trends in Mathematics and Science Study (TIMSS) dataset that is a large-scale, international, nested, secondary dataset. This study seeks to determine what model class (random forest, gradient-boosted trees, multivariate adaptive regression splines, or stacked generalization) is best at reducing model error in predicting student math achievement and how these models differ across 39 countries. By using cross-validated iterative ML techniques, it also aims to establish the student, teacher, and school characteristics that are critical in predicting math achievement among 8th grade students. While the methods in this study do not use inferential statistics to examine math achievement, the predictive modeling techniques utilized may help us shed light on the contextual factors and/or culture that may account for differences in student math achievement as well as how analogous these modeled traits are across countries.
Add important information, links, or images here to describe your project.