Using Machine Learning Algorithms to Classify Fire Risk in the Municipality Of Altamira in The State of Pará

Forest fires are a serious problem in Brazil. According to data from the National Institute for Space Research (INPE), the state of Pará, in Brazil, is home to the two most affected municipalities in the period of 2024, São Félix do Xingu and Altamira, causing environmental and health damage to the...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:2025 IEEE Latin Conference on IoT (LCIoT) S. 111 - 115
Hauptverfasser: Lopes da Costa, Nicole Victoria, Juvenal Menezes Filho, Averaldo, Santos Parente, Juan Carlos, Martins de Souza, Laiane, Di Paolo, Italo Flexa, de Oliveira Barreiros, Marta
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 23.04.2025
Schlagworte:
Algorithms
Altamira
Classification
Classification algorithms
Fauna
Fire Risk
Forestry
Machine Learning
Machine learning algorithms
Monitoring
Nearest neighbor methods
Pará
Prediction algorithms
Prevention and mitigation
Support vector machines
Wildfires
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Forest fires are a serious problem in Brazil. According to data from the National Institute for Space Research (INPE), the state of Pará, in Brazil, is home to the two most affected municipalities in the period of 2024, São Félix do Xingu and Altamira, causing environmental and health damage to the population, highlighting the need for studies on the subject. Therefore, we analyzed the municipality of Altamira to study the classification of forest fire risk, between low, medium or high, correlating meteorological data from the National Institute of Meteorology (INMET) with fire records from INPE. In this study, we used the K-Nearest Neighbors (KNN), Extra Trees and Support Vector Machines (SVM) algorithms to classify forest fire risk, and the results show that the Extra Trees algorithm obtained the best performance with 84% accuracy, followed by KNN with 80.2% and SVM with 78.4%.
DOI:10.1109/LCIoT64881.2025.11118543