A Comparison of One-Class Versus Two-Class Machine Learning Models for Wildfire Prediction in California

Fathima Nuzla Ismail, Abira Sengupta, Brendon J. Woodford, Sherlock A. Licorish

Abstract: Due to climate change, forest regions in California are increasingly experiencing severe wildfires, with other issues affecting the rest of the world. Machine learning (ML) and artificial intelligence (AI) models have emerged to predict wildfire hazards and aid mitigation efforts. However, the wildfire prediction modelling domain faces inconsistencies due to database manipulations for multi-class classification. To help to address this issue, our paper focuses on creating wildfire prediction models through One-class classification algorithms: Support Vector Machine, Isolation Forest, AutoEncoder, Variational AutoEncoder, Deep Support Vector Data Description, and Adversarially Learned Anomaly Detection. To minimise bias in the selection of the training and testing data, Five-Fold Cross-Validation was used to validate all One-class ML models. These One-class ML models outperformed Two-class ML models using the same ground truth data, with mean accuracy levels between 90 and 99 percent. Shapley values were used to derive the most important features affecting the wildfire prediction model, which is a novel contribution to the field of wildfire prediction. Among the most important factors were the seasonal maximum and mean dew point temperatures. In providing access to our algorithms, using Python Flask and a web-based tool, the top-performing models were operationalized for deployment as a REST API, with the potential to strengthen wildfires mitigation strategies.