Detecting Key Indicators of Student Concentration in Mathematics Using TIMSS 2023 Data: An Explainable Machine Learning Approach

Abstract

In Morocco there is an enormous issue to keep students engaged and focused on a math class because this directly affects the way they learn and perform at schools overall. A 2022 to 26 June Strategic Roadmap for Education Reform clarified that the loss of concentration is a scourge that affects performance and participation. With the help of data from the TIMSS 2023 assessment, which focuses on students in the eighth grade, this study seeks to identify the major factors that affect the ability of students to stay focused in mathematics classes. Prior to analysis, we cleansed and processed the data by eliminating the absence of entries and by aggregating related categories. After that, we used an autoencoder to reduce the size of the data and help to reveal the hidden patterns in behavior and contexts in the classroom. Based on these capabilities, several models, such as random forests, decision trees, XGBoost, and an unstructured neural network, were developed to categorize students based on their attention level as either low or high. SHAP values revealed which variables were the most influential on the students' attention, making it easier to understand the findings. The results highlighted some persistent factors: the environment of the classroom and instructional delivery, technology accessibility, and the quality of interactions between teachers and students are all essential for maintaining focus. The research was able to show that explicable AI can be an extremely asset in research on education, providing explicit information and strategies that can be implemented by teachers and policymakers in the position to facilitate involvement in mathematics education.