modeling The possibility of changing urban growth using artificial neural network and logistic regression (Case Study: Mashhad)

In developing countries, the high tendency for concentration of population in urban areas and consequently the rapid and uneven growth of cities have led urban designers and planners to use appropriate policies and strategies to avoid environmental and socio-economic damaging effects on the order Work. In this regard, spatial and temporal information related to growth rate patterns provides a better understanding of the urban growth process and provides appropriate tools for obtaining management and planning policies for urban managers. Therefore, the main objective of this research is to calculate the probability of growth change in Mashhad using logistic regression and artificial neural network. For this purpose, satellite images of Landsat 7 (2002) and Landsat 8 (2015) were used to provide land-use mapping. Then, using multi-layer perceptron artificial neural network (MLP), the classification of images was made and urban land use maps with a total accuracy of 948/0 and a Kappa index of 936 for 2002 as well as a general accuracy of 8177 and a Kappa index of 775 / 0 were extracted for 2015. Finally, with the implementation of logistic regression between urban land use map 2015 (as dependent variable) and effective factors such as physical factors and human factors along with 2002 map of lands (as independent variables), the potential map of urban land development was prepared. The evaluation of the regression model generated by two Pseudo-R2 and ROC indicators showed that this model has a ROC value of 0.87 and Pseudo-R2 of 345/0 has a high ability to show changes and determine areas prone to change, and fit The model is considered fairly well.