Forest Fire Prediction using Random Forest

Forest fires pose significant ecological, economic, and human threats, demanding accurate and timely prediction systems for effective mitigation. This study proposes a machine learning-based approach using the Random Forest (RF) algorithm to predict the likelihood of forest fire occurrences based on environmental and meteorological variables. Utilizing the UCI Forest Fires dataset, the model classifies fire risk by analyzing features such as temperature, relative humidity, wind speed, drought code, and the initial spread index. The dataset underwent preprocessing steps including normalization, label binarization, and feature encoding to ensure consistency and model readiness. The Random Forest model achieved a high accuracy of 93.6%, outperforming conventional classifiers such as SVM, ANN, KNN, Naive Bayes, and Decision Trees. Evaluation metrics including precision (0.94), recall (0.92), F1-score (0.93), and ROC-AUC (0.96) further affirm its robust predictive performance. Feature importance analysis demonstrated alignment with established fire risk indicators, improving the interpretability of the model. Compared to five peer-reviewed approaches, this study presents superior classification capabilities and offers a scalable, reliable solution for forest fire management. Future work will focus on expanding the dataset, integrating spatial-temporal and satellite data, and deploying hybrid and explainable AI models to enhance operational readiness in real-world applications.