作者: Omid Ghorbanzadeh , Hashem Rostamzadeh , Thomas Blaschke , Khalil Gholaminia , Jagannath Aryal
DOI: 10.1007/S11069-018-3449-Y
关键词: Geographic information system 、 Groundwater 、 Adaptive neuro fuzzy inference system 、 Natural hazard 、 Cross-validation 、 Data mining 、 Receiver operating characteristic 、 Hydrogeology 、 Subsidence 、 Geology
摘要: In this paper, we evaluate the predictive performance of an adaptive neuro-fuzzy inference system (ANFIS) using six different membership functions (MF). combination with a geographic information (GIS), ANFIS was used for land subsidence susceptibility mapping (LSSM) in Marand plain, northwest Iran. This area is prone to droughts and low groundwater levels subsequent damages. Therefore, inventory database created from extensive field survey. Areas or areas showing initial signs were training, while onethird reserved testing validation. The randomly divided into three folds same size. One chosen Other two training. process repeated every fold dataset. Thereafter, related factors, such as hydrological topographical prepared GIS layers. susceptible then analyzed approach, maps created, whereby MFs applied. Lastly, results derived each MF validated those that not Receiver operating characteristics (ROC) curves drawn all LSSMs, under calculated. ROC analyses LSSMs yielded very high prediction values out methods, namely difference DsigMF (0.958) GaussMF (0.951). integration generally led LSSM accuracies. study demonstrated choice training dataset significantly affects results.