TY - JOUR
T1 - Hybrid artificial intelligence models based on a neuro-fuzzy system and metaheuristic optimization algorithms for spatial prediction of wildfire probability
AU - Jaafari, Abolfazl
AU - Zenner, Eric K.
AU - Panahi, Mahdi
AU - Shahabi, Himan
N1 - Publisher Copyright:
© 2018 Elsevier B.V.
PY - 2019/3/15
Y1 - 2019/3/15
N2 - This study provides a new comparative analysis of four hybrid artificial intelligence models for the spatially explicit prediction of wildfire probabilities. Each model consists of an adaptive neuro-fuzzy inference system (ANFIS) combined with a metaheuristic optimization algorithm, i.e., genetic algorithm (GA), particle swarm optimization (PSO), shuffled frog leaping algorithm (SFLA), and imperialist competitive algorithm (ICA). A spatial database was constructed based on 159 fire events from the Hyrcanian ecoregion (Iran) for which a suite of predictor variables was derived. Each predictor variable was discretized into classes. The step-wise weight assessment ratio analysis (SWARA) procedure was used to assign weights to each class of each predictor variable. Weights indicate the strength of the spatial relationship between each class and fire occurrence and were used for training the hybrid models. The hybrid models were validated using several performance metrics and compared to the single ANFIS model. Although the single ANFIS model outperformed the hybrid models in the training phase, its accuracy decreased considerably in the validation phase. All hybrid models performed well for both training and validation datasets, but the ANFIS-ICA hybrid showed superior predictive performance of spatially explicit wildfire prediction and mapping for the dataset. The results clearly demonstrate the ability of the optimization algorithms to overcome the over-fitting problem of the single ANFIS model at the learning stage of the fire pattern. This study contributes to the suite of research that seeks to obtain reliable estimates of relative likelihoods of natural hazards.
AB - This study provides a new comparative analysis of four hybrid artificial intelligence models for the spatially explicit prediction of wildfire probabilities. Each model consists of an adaptive neuro-fuzzy inference system (ANFIS) combined with a metaheuristic optimization algorithm, i.e., genetic algorithm (GA), particle swarm optimization (PSO), shuffled frog leaping algorithm (SFLA), and imperialist competitive algorithm (ICA). A spatial database was constructed based on 159 fire events from the Hyrcanian ecoregion (Iran) for which a suite of predictor variables was derived. Each predictor variable was discretized into classes. The step-wise weight assessment ratio analysis (SWARA) procedure was used to assign weights to each class of each predictor variable. Weights indicate the strength of the spatial relationship between each class and fire occurrence and were used for training the hybrid models. The hybrid models were validated using several performance metrics and compared to the single ANFIS model. Although the single ANFIS model outperformed the hybrid models in the training phase, its accuracy decreased considerably in the validation phase. All hybrid models performed well for both training and validation datasets, but the ANFIS-ICA hybrid showed superior predictive performance of spatially explicit wildfire prediction and mapping for the dataset. The results clearly demonstrate the ability of the optimization algorithms to overcome the over-fitting problem of the single ANFIS model at the learning stage of the fire pattern. This study contributes to the suite of research that seeks to obtain reliable estimates of relative likelihoods of natural hazards.
UR - http://www.scopus.com/inward/record.url?scp=85059424240&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85059424240&partnerID=8YFLogxK
U2 - 10.1016/j.agrformet.2018.12.015
DO - 10.1016/j.agrformet.2018.12.015
M3 - Article
AN - SCOPUS:85059424240
SN - 0168-1923
VL - 266-267
SP - 198
EP - 207
JO - Agricultural and Forest Meteorology
JF - Agricultural and Forest Meteorology
ER -