Comparação de algoritmos preditivos para incêndios em canaviais

Abstract

The occurrence of fire in sugarcane fields is a problem that has plagued several regions of the country for a long time. Due to the concern about this and its tragic consequences, it is necessary to take actions in order to avoid or alleviate this problem. With the development of Artificial Intelligence (AI) and the growing use of Machine Learning (ML), there is an opportunity to use technology in favor of anticipating an imminent fire. In this context, the objective of this work was to implement predictive algorithms using the Python language to compare and define what best applies to fire prediction in sugarcane fields. The data necessary for this purpose were provided by a sugarcane company located in the interior of São Paulo. Among the four analyzed algorithms, these being the Support Vector Machine (SVM), Naive Bayes, Random Forest and XGBoost, the SVM model showed a better performance against the performance metrics used.