Machine Learning for Early Fire Detection in the Oasis Environment

An oasis ecosystem that consists of date palms prevents desertification and ensures the survival of a rich biodiversity. However, climate change is having a negative impact on our planet by increasing global temperatures, leading to fires in the oasis environment and, consequently, threatening the activity of the population. In this paper, we aim to build an accurate model to detect fire extinguishers as early as possible in the oasis area. To this end, we tested several machine learning algorithms, namely AdaBoost, Naïve Bayes, Random Forest, kNN, Gradient Boosting, and Neural Network applied to a dataset containing 1900 images in total. The results show that the Neural Network delivers the highest accuracy at 97.7%.