Abstract
There has been a tremendous increase in recorded forest fires around the globe. In California itself, nearly 85 million acres of land are covered with wildfires. In 2020, a total of 10 million acres was destroyed due to wildfires. . The forest fires can be of different forms and sizes depending on the weather conditions, the characteristic of the vegetation, and landcover type. So, to better manage it and to reduce human, economic, and environmental loss, we can utilize machine learning techniques to design an efficient forest fire prediction system. Forest fire prediction can play a huge role in controlling the fires and reducing the destruction caused due to them. The design of an efficient forest fire prediction can help in reducing the destruction caused by these fires in the future. To predict the event of the forest fire the proposed approach undertakes different meteorological parameters i.e. temperature, rain, wind, and humidity, and landcover type to identify the high-risk areas where fires can start. In this project, several machine learning algorithms are used to predict the high-risk areas using previous forest fires data in different scenarios. Considering the performance of different models with different features to find the best way to predict the occurrence and spread of fires. After doing the experiments, the best model is chosen with the features which help in predicting the occurrence and project their intensity even before the fire has been started.