International Journal of Electrical and Computer Engineering (IJECE) Vol. 14, No. 2, April 2024, pp. 1842~1850 ISSN: 2088-8708, DOI: 10.11591/ijece.v14i2.pp1842-1850  1842 Journal homepage: http://ijece.iaescore.com Predicting and detecting fires on multispectral images using machine learning methods Murat Aitimov 1 , Mira Kaldarova 2 , Akmaral Kassymova 3 , Kaiyrbek Makulov 4 , Raikhan Muratkhan 5 , Serik Nurakynov 6 , Nurmakhambet Sydyk 6 , Ideyat Bapiyev 3 1 Kyzylorda Regional Branch at the Academy of Public Administration under the President of the Republic of Kazakhstan, Kyzylorda, Republic of Kazakhstan 2 Department of Information Systems, S. Seifullin Кazakh Research Agrotechnical University, Astana, Republic of Kazakhstan 3 Department of Information Technology, Faculty of Technology, Zhangir Khan University, Uralsk, Republic of Kazakhstan 4 Department of Computer Science, Faculty of Science and Technology, Caspian University of Technology and Engineering Named after Sh. Yessenov, Aktau, Republic of Kazakhstan 5 Department of Applied mathematics and Informatica of Karaganda Buketov University, Karaganda, Republic of Kazakhstan 6 Institute of Ionosphere, Almaty, Republic of Kazakhstan Article Info ABSTRACT Article history: Received Aug 27, 2023 Revised Oct 24, 2023 Accepted Nov 29, 2023 In today's world, fire forecasting and early detection play a critical role in preventing disasters and minimizing damage to the environment and human settlements. The main goal of the study is the development and testing of machine learning algorithms for automated detection of the initial stages of fires based on the analysis of multispectral images. Within the framework of this study, the capabilities of three popular machine learning methods: extreme gradient boosting, logistic regression, and vanilla convolutional neural network (vanilla CNN), are considered in the task of processing and interpreting multispectral images to predict and detect fires. XGBoost, as a gradient-boosted decision tree algorithm, provides high processing speed and accuracy, logistic regression stands out for its simplicity and interpretability, while vanilla CNN uses the power of deep learning to analyze spatial and spectral data. The results of the study show that the integration of these methods into monitoring systems can significantly improve the efficiency of early fire detection, as well as help in predicting potential fires. Keywords: Extreme gradient boosting Fire Logistic regression Machine learning Multispectral images Vanilla convolutional neural network This is an open access article under the CC BY-SA license. Corresponding Author: Ideyat Bapiyev Department of Information Technology, Faculty of Technology, Zhangir Khan University 090000 Uralsk, Republic of Kazakhstan Email: bapiev@mail.ru 1. INTRODUCTION Forest fires are one of the most destructive natural phenomena, causing significant damage to ecosystems, economies, and human well-being. In light of global climate change and increasing human impacts, the need for rapid and accurate fire prediction and detection has become increasingly urgent. The main challenge here is the need for effective fire prediction and detection in multispectral images. To solve this problem, it is proposed to develop and optimize highly efficient machine learning models capable of analyzing multispectral images to accurately predict and detect fires. These models must be trained on large amounts of data, including a variety of scenarios and types of multispectral data, to ensure their broad applicability. Traditional methods of detection and control often do not provide the necessary effectiveness, so the focus is on innovative approaches. In recent decades, the development of Earth remote sensing technologies has made it possible to obtain high-resolution multispectral images. These images [1]–[3]