خلاصة:
Artificial neural network models are considered information-based models. The land use/land cover change model is one of the models that connects artificial neural networks with geographic information systems and was used to model urban development in Gonbade Kavus during the period 2001-1987. This model consists of 6 applications that run on the MS-DOS program. In this study, three groups of variables including biophysical, socio-economic and land use variables were used. Ten variables affecting urban development from the aforementioned three groups were used as input nodes and the dependent variable of urban changes as output nodes. Using hidden nodes was for identifying non-linear relationships in the model. Performing the model in the 3000th cycle had the lowest amount of mean squared error; therefore, this cycle was used to extract future urban development areas and model sensitivity analysis. To validate the model, the probability image of this cycle was used, for which the Relative Operating Characteristic (ROC) statistical approach was estimated at 0.75, which validated the model under the aforementioned conditions. Using this image, the pattern of urban distribution was extracted for the years 2010, 2020, 2030 and 2040. The relative effect of variables was estimated using ROC and by removing one variable and running the model with the remaining variables and comparing it with the complete model. To do this, 11 networks were created with complete data and data without one variable, and the training and testing stages were executed for all of them. Land use of cultivated lands, the number of urban cells, and rangeland use had the most effect during the study period; and the distance to educational centers, forest use, and barren lands had the least effect on the growth of Gonbade Kavus city; in other words, it was found that the type of land use in the area has a significant impact on urban development in Gonbade Kavus.
ملخص الجهاز:
Application of Artificial Neural Network in Urban Development Modeling (Case Study: City of Gorgan) Hamid Reza Kamiyab{o*o} - Master of Science in Natural Resources Engineering - Environment, Tarbiat Modares University, Tehran, Iran.
The land use/cover change model is one of the models that connects artificial neural networks with geographic information systems and was used to model urban development in the city of Gorgan during the period 1987-2001.
This model, which connects the Geographic Information System{o2o} with an artificial neural network, has the ability to import a wide variety of inputs effective on land use changes such as socio-economic, environmental, and political factors.
)COR(citsiretcarahC gnitarepO evitaleR o} was prepared, along with the variable of urban changes in the study area during the period 2001-1987 for use in the model.
The variables affecting the development of Gorgan city, which are considered as the nodes of the MTL model, include input nodes (variables affecting urban development) and output nodes (urban changes during the study period).
Model Sensitivity Analysis To express the amount of influence of each of the variables used on urban changes in the 2001-1987 time period, the method of removing one variable and running the model with the remaining data and examining the goodness of fit of the model with COR is used )3002,renruT & ynihaM( .
In this study, regression modeling and neural network methods were used to identify and improve our understanding of the socio-economic, physical, and land use forces that affect urban development, as well as to find the most likely locations for future urban development in Gorgan.
(
, 
, 
)
