Integration of ARIMA Models and Machine Learning for Academic Data Forecasting - A Case Study in Applied Mathematics

Aug 3, 2026

This study explores ARIMA models and machine learning algorithms, specifically Random Forest and Multiple Linear Regression, for predicting student academic performance. A mixed-method approach analyzes three years of academic grade data, using ARIMA to capture time-series trends and using machine learning to model outcomes from multiple explanatory variables. The results indicate that ARIMA effectively maps academic trend dynamics, while Random Forest performs strongly in representing complex variable relationships. Reported errors include RMSE 1.12 and MAE 0.94, supporting data-driven educational decision-making.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages8

Reading Time20 min

AccessFree to download

Abstract

Introduction

Challenges in applied mathematics learning

Limitations of traditional teacher judgment

Rationale for ARIMA + machine learning

Related Work

ARIMA and statistical/probabilistic models in education

Machine learning approaches for academic prediction

Zero : Jurnal Sains, Matematika, dan Terapan E-ISSN : 2580-5754; P-ISSN : 2580-569X Volume 9, Number 1, 2025 DOI: 10.30829/zero.v9i1 .24148 Page: 297-304 | Integration of ARIMA Models and Machine Learning for Academic Data Forecasting: A Case Study in Applied Mathematics
1 Erwinsyah Simanungkalit
Politeknik Negeri Medan, Indonesia
2 Mardhiatul Husna
Politeknik Negeri Medan, Indonesia
3 Jenny Sari Tarigan
Politeknik Negeri Medan, Indonesia | | | --- | --- | | Article Info
Article history:
Accepted, 30 July 2025
Keywords:
Academic Data; ARIMA;
Machine Learning; Predictive Analytics;
Corresponding Author: | ABSTRACT
This study explores the use of ARIMA models and machine learning algorithms, specifically Random Forest and Multiple Linear Regression, to predict student academic performance. A mixed-method approach analyzed academic grades data from the past three years, with ARIMA identifying time series trends and machine learning models predicting academic outcomes based on various variables. Results show ARIMA effectively maps academic trends, while Random Forest excels in handling complex relationships, with an RMSE of 1.12 and an MAE of 0.94. These findings highlight the potential of combining statistical models and machine learning in developing adaptive learning strategies and data-driven decision-making. This approach offers a robust framework for improving educational outcomes and can guide future research in predictive analytics for educational systems.
This isan open access article under the CC BY-SA license.
| | Erwinsyah Simanungkalit,
Department of Business Management Politeknik Negeri Medan, Indonesia, Email: [erwinsyahsimanungkalit@polmed.ac.id](erwinsyahsimanungkalit@polmed.ac.id) | | | 1. INTRODUCTION
Applied mathematics plays an important role in modern education, especially in connecting theoretical concepts with practical applications in various fields. However, many students do not yet know its application and often face challenges in understanding and applying applied mathematics concepts effectively. Difficulties occur due to various internal and external factors such as the use of teaching approaches that still emphasize too much on memorizing formulas and procedures without providing a deep understanding of the real applications of these concepts [1] .
Several studies have identified challenges in applied mathematics learning. Developed an intervention in mathematics learning and education to help students who are struggling and at risk of having difficulty understanding mathematics, showing that a tailored approach can improve students' understanding. In addition, an | | Journal homepage:[http://jurnal.uinsu.ac.id/index.php/zero/index](http://jurnal.uinsu.ac.id/index.php/zero/index) approach to mathematics learning that still emphasizes exploration and understanding rather than performance has been shown to improve students' motivation and learning outcomes. One of the main problems in traditional teacher practice is the reliance on subjective judgment and personal experience in teaching and predicting student academic performance, which is often not supported by historical or objective data analysis. This approach is prone to bias, is less adaptive to the complexity of the modern education system, and has the potential to cause delays in intervention for students with learning difficulties. In this context, the integration of ARIMA models and Machine Learning algorithms becomes very relevant as a solution based on Applied Mathematics. ARIMA can identify long-term trends in academic score data, while algorithms such as Random Forest are able to capture the complex relationships between variables that influence student performance. The combination of these two methods allows for more accurate, adaptive, and evidence-based decision-making, so that teachers can design more targeted interventions and support more responsive learning st

FAQs

Which models are used to forecast student academic performance?

The study uses ARIMA for time-series trend identification and two machine learning methods—Random Forest and Multiple Linear Regression—to predict outcomes based on multiple variables.

How was the data and analysis approach conducted?

A mixed-method approach analyzes academic grades collected over the past three years, combining ARIMA trend modeling with machine learning outcome prediction.

What do the results show about ARIMA and Random Forest performance?

ARIMA is effective at mapping academic trends, while Random Forest better captures complex relationships among variables; the reported performance includes RMSE 1.12 and MAE 0.94.

Aug 3, 2026

Upload by Seraphina