The use of maximum entropy and ecological niche factor analysis to decrease uncertainties in samples for urban gain models

Uncertainty is a common problem in spatial modeling and geographical information systems (GIS). Furthermore, urban gain (UGM) contains various dimensions components of uncertainties. Data sampling important UGM, may cause the results models to contain many uncertainties as well affects their precision accuracy. A poorly sampled or biased dataset can lead inaccurate predictions decreased performance models. This paper aims present develop novel strategies for building training datasets that enhance data-driven In other words, study used maximum entropy (ME) ecological niche factor analysis (ENFA) select pure non-change samples with minimal uncertainty UGM Isfahan Tabriz cities Iran. The two time intervals 1992–2002 2002–2012 were City 1994–2004 2004–2014 City. Nine 14 drivers cities, respectively. After ME ENFA produced change lowest uncertainty, three well-known models, namely random forest (RF), artificial neural network (ANN), support vector machine (SVM) modeling. Moreover, investigate procedure one-class prediction Compared extant studies, proposed – based strategy increased area under receiver operating characteristic curve (AUROC), figure merit, producer’s accuracy, overall accuracy by 5.5%, 5%, 3%, respectively, validation phase 6%, 14%, 17%, For Isfahan, accuracies (AUROC = 0.649) 0.661) one class closer ANN 0.646), 0.619), RF 0.631) but differed significantly from 0.737) model. Tabriz, 0.657) 0.688) lower than RF-ME 0.852), ANN-ME 0.778) showed model was able identify relatively properly remove impure dataset. discovered binary are preferable an optimal essential step it decrease uncertainty. As such, modelers must adopt efficient methods.