Gene expression programming and data mining methods for bushfire susceptibility mapping in New South Wales, Australia

Abstract Australia is one of the most bushfire-prone countries. Prediction and management bushfires in bushfire-susceptible areas can reduce negative impacts bushfires. The generation bushfire susceptibility maps help improve prediction main aim this study was to use single gene expression programming (GEP) ensemble GEP with well-known data mining generate for New South Wales, Australia, as a case study. We used eight methods mapping: GEP, random forest (RF), support vector machine (SVM), frequency ratio (FR), techniques FR (GEPFR), RF (RFFR), SVM (SVMFR), logistic regression (LR) (LRFR). Areas under curve (AUCs) receiver operating characteristic were evaluate proposed methods. GEPFR exhibited best performance mapping based on AUC (0.892 training, 0.890 testing), while RFFR had highest accuracy (95.29% 94.70% testing) among an method that uses features from evolutionary algorithm statistical method, which results better maps. Single showed 0.884 training 0.882 testing. also 0.902 0.876 testing, respectively. 0.868 0.781 testing mapping. performances than those