PREDICTING MATHEMATICS LEARNING BEHAVIOR IN ELEMENTARY SCHOOL STUDENTS: A MACHINE LEARNING PERSPECTIVE

Abstrak

In the digital era, understanding students' learning behavior is crucial for improving the quality of education. This research employs machine learning algorithms to predict elementary students' mathematics learning behavior based on their learning activity data, such as problem-solving time, success rate, and interaction patterns in technology-based learning. By leveraging the Random Forest algorithm, this study successfully identifies key factors influencing students' learning behavior.

The study aims to predict the learning behavior of elementary students using the Random Forest algorithm, implemented on 6th-grade students from Class 6-B, SD Negeri Baru 01 Pagi Jakarta. Features including the number of problems solved, average solving time, success rate, and repetition frequency were utilized to build the prediction model. The results revealed that the model achieved 100% accuracy on test data. The developed Random Forest model demonstrated excellent performance in predicting students' learning behavior based on the available dataset features, with detailed findings as follows: (1) Average Solving Time contributed the most (35.78%) to the model's predictions; (2) Success Rate contributed 32.57%, indicating that students' success in solving problems is a significant factor in predicting learning behavior; (3) Repetition Frequency accounted for 20.46%, reflecting how students' persistence in repeating exercises impacts learning behavior predictions; and (4) Number of Problems Solved had the lowest contribution (11.19%), suggesting that this feature has a smaller impact compared to others.

The findings provide valuable insights for educators to design more effective learning strategies, including personalized interventions for students at risk of low motivation.