APPLICATION OF ADAPTIVE NEURO-FUZZY INTERFERENCE SYSTEM MODELS FOR PREDICTION OF FOREST FIRES IN THE USA ON THE BASIS OF SOLAR ACTIVITY

摘要

In this research we search for a functional dependence between the occurrence of forest fires in the USA and the factors which characterize the solar activity. For this purpose we used several methods (R/S analysis, Hurst index) to establish potential links between the influx of some parameters from the sun and the occurrence of forest fires with lag of several days. We found evidence for a connection and developed a prognostic scenario based on the Adaptive neuro-fuzzy interference system (ANFIS) technique. This scenario allows the prediction between 79-93% of forest fires. [Projekat Ministarstva nauke Republike Srbije, br. III47007]