Learning Difficulty Levels Prediction of Elementary School Student Mathematics Using Machine Learning Model

Rismayani Rismayani (Universitas Dipa Makassar)
Novita Sambo Layuk (Universitas Dipa Makassar)
Madyana Patasik (Universitas Dipa Makassar)
Andi Hutami Endang (Institut Teknologi dan Bisnis Kalla, Indonesia)

Journal of Information Technology and its Utilization
Vol 8 No 1 (2025): June 2025

Difficulty learning mathematics in elementary school students is a significant problem and requires serious attention. This study aims to predict the difficulty level in elementary school students learning mathematics using a machine learning model, namely KNN. Exam scores, assignments, quizzes, and characteristics of students' difficulty level in learning mathematics were used as data in this study. A study used the KNN model to divide students into three categories of difficulty in learning mathematics: easy, moderate, and challenging. The results showed that the KNN model can accurately predict student’s difficulty levels in mathematics. Thus, applying this model can help teachers provide appropriate and effective interventions to students experiencing difficulties. Using machine learning technology, especially the KNN model, we found an accuracy of 95%. In addition, we can still accurately predict the difficulty level of elementary school students' mathematics learning. This study uses anonymous student data, the distribution of assignments, quizzes, and exam score ranges, and characteristics of mathematics learning difficulty levels. There are three prediction classes: high, medium, and low.