TIMSS 2011 Predictors Relating to Korean 8th Graders’ Mathematics Achievement, Explored Via Machine Learning

Source abstract

A substantial body of research has been conducted on factors relating to students’ math achievement with TIMSS. However, most studies have focused on selected a few factors instead of utilizing hundreds of variables TIMSS provides, and have employed conventional statistical methods. This study aimed to investigate possible sets of predictors from a totally different approach: LASSO, currently one of the most popular machine learning techniques. Korean 8th graders’ TIMSS 2011 were used as the sample, and the prediction accuracy of the LASSO model was about 80% with the selected 22 out of 100 predictors. As results, students’ math efficacy, attitudes toward math, mother’s education level, and home educational resources including amount of books at home were influential to their math achievement, which was consistent with previous studies. Additionally, math homework completion time, student’s science self-efficacy, and science homework frequency were newly found important predictors. Implications and future research topics are discussed.