Fusion of socio-economic and remote sensing-based attributes for Karaj physical growth modeling

Urban physical growth is affected by different parameters including environmental, neighborhood and socio-economic factors; however, socio-economic variables are often ignored due to the lack of socio-economic information, especially in developing countries, when the urban physical growth analysis and modeling is the aim. Accordingly, there is not many studies conducted to develop GIS-based socio-economic layers to be used along with common data, such as slope, distance to the roads and so on, in urban physical growth modeling. Therefore, this study aims to introduce an efficient method to generate GIS-based socio-economic layers to be exploited along with the information layers extracted from Landsat images and field-collected data for physical growth modeling of Karaj city. After generating the required information layers, random forest feature selection method was applied to select the most important variables. Then, the performance of the three modeling methods including multiple logistic regression, and two artificial neural networks, multi-layer perceptron (MLP) and self-organizing map (SOM) were compared using the selected attributes to model the urban physical growth from 2000 to 2010. The results indicated that SOM with overall accuracy of 84.5%, kappa coefficient of 68.9%, ROC of 90.7%, FOM of 43.98% and PCM of 84.5% performed better than the other methods for modelling of urban physical growth. Moreover, the proposed socio-economic attributes combined with the remote sensing-based data were able to improve the performance of the urban physical growth prediction. Finally, cellular automata was applied to predict the Karaj physical growth in 2017 and 2027.