Early Detection of Mathematics Learning Difficulties Using a Deep Learning Neural Network and SMOTE

DOI

10.2991/978-94-6463-924-7_21How to use a DOI?

Keywords

Mathematics Learning Difficulty Prediction; Deep Learning Neural Network; CRISP-DM; Educational Early Warning System; Educational Decision Support System

Abstract

This study follows the CRISP-DM (Cross-Industry Standard Process for Data Mining) framework, a structured methodology that guides data mining projects through six stages: business understanding, data under-standing, data preparation, modeling, evaluation, and deployment. The dataset used in this research consists of academic variables (grades, remedial sessions, online activities) and psychological variables (motivation, anxiety, interest) collected from SMA Muhammadiyah 16 Jakarta. The workflow in-volves data preprocessing, data augmentation, class balancing using SMOTE, and training a neural network model with four hidden layers using ReLU and Softmax activation functions, focal loss, and class weight balancing. The model achieved an accuracy of 98%, along with an average recall and F1-score of 0.98, indicating excellent performance across all classes. Further analysis showed that semester exam scores, number of remedial sessions, motivation levels, and self-confidence were the most influential factors in distinguishing the categories “easy,” “moderate,” and “difficult.” These findings demonstrate that the model effectively identifies patterns of learning difficulties and can support teachers in providing targeted interventions. Thus, the research objective of developing a highly accurate predictive mod-el has been successfully achieved. Moreover, this study makes a significant contribution to the development of AI-based decision support systems in education and introduces a prototype of a web-based early warning system for schools.

Copyright

© 2025 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Harry Dhika
AU  - Surajiyo Surajiyo
AU  - Lasia Agustina
AU  - Abdul Muchlis
PY  - 2025
DA  - 2025/12/16
TI  - Early Detection of Mathematics Learning Difficulties Using a Deep Learning Neural Network and SMOTE
BT  - Proceedings of the 8th International Conference on Informatics, Engineering, Science & Technology (INCITEST 2025)
PB  - Atlantis Press
SP  - 251
EP  - 268
SN  - 2352-5401
UR  - https://doi.org/10.2991/978-94-6463-924-7_21
DO  - 10.2991/978-94-6463-924-7_21
ID  - Dhika2025
ER  -