Integrating XAI and Machine Learning for an Effective Forest Fire Prediction System
Abstract
One of the most damaging natural catastrophes, forest fires greatly increase global carbon emissions while wreaking damage on ecosystems, biodiversity, and human communities. Implementing timely preventative methods, allocating resources optimally, and minimizing ecological and financial losses all depend on early and accurate forest fire forecasts. This study introduces a machine learning-based method for forest fire prediction that makes use of logistic regression. We use the Synthetic Minority Over-sampling Technique (SMOTE) to rectify the dataset’s class imbalance and guarantee a balanced representation of fire and non-fire events throughout training. Additionally, we visualize data using Principal Component Analysis (PCA), which sheds light on how separable classes are within the feature space. Our model’s resilience in categorizing forest fire incidents is demonstrated by its outstanding accuracy and F1-score of 98%. The suggested method is appropriate for real-time implementation in forest management systems as it provides a simple yet efficient solution.