Back in 2012, an estimated 20,000 hectares of land have been ravaged by forest fires in the north of Algeria between June and September. Forest fire is a disaster that leads to serious issues to the affected nation which causes increasing carbon, climate change and threatening the humanity as a whole. So predicting such essential environmental issue is essential to mitigate this threat. during this paper the main purpose is to use Data Science (DS) and Machine Learning models (ML) to predict forest fire outbreak which are presented on this paper and they depend on weather elements and Fire Weather Index (FWI) components.

The main idea is proposing a machine learning models and fire prediction based on integrating some data mining techniques that classify whether the region is on or not on fire based on certain attributes, for that and since the classification is two classes (Fire & Not Fire), the binary classification was considered. As a first step, performing some analysis on the data by applying some EDA along with some data preprocessing to make the data ready to be fitted into a Machine Learning model as well as creating some data visuals for a better understanding of the dataset, for the second step creating some ML predictive models for the binary classification and compare between their different performance and choosing the best one in order to create a classification report for the chosen model. the last section was devoted to the clustering of the dataset and interpreting the clusters into low-moderate-high and very high danger fire using the proposed graph of Forest Fire Danger Class Criteria from the National Rural Fire Authority.