An enhanced support vector machine model for urban expansion prediction. (May 2019)

• Record Type: Journal Article
• Authors: Karimi, Firoozeh; Sultana, Selima; Shirzadi Babakan, Ali; Suthaharan, Shan
• Is Part Of: Computers, environment and urban systems. Volume 75(2019)
• Page Start: 61
• Page End: 75
• Publication Date: 2019-05
• Subjects: Urban growth -- GIS -- Machine learning -- Support vector machine (SVM) -- Land-use/land-cover change; City planning -- Data processing -- Periodicals; Regional planning -- Data processing -- Periodicals; 303.4834
• DOI: 10.1016/j.compenvurbsys.2019.01.001
• Languages: English
• ISSNs: 0198-9715

Abstract

Abstract: Given the dynamic and convoluted nature of urban expansion process and the necessity of handling continuous and categorical variables, non-normal distributed data, and non-linear relationships, urban expansion modeling is challenging. To handle these issues effectively and enhance the quality of urban expansion prediction, the capabilities of support vector machine (SVM) technique are explored in this study. A binary SVM model is developed using three different data sampling methods and nineteen predictor variables, four of which are first introduced in this study. The model is configured by regulating the penalty parameter, selecting the most appropriate kernel function, and setting the best value for the kernel function's parameter. A novel combination of goodness-of-fit metrics is used to more realistically evaluate the model accuracy to predict built and unbuilt land cells as well as changed and unchanged land cells in the whole study area. The implementation of the developed model in Guilford County, NC, over the period of 2001–2011, as a case study, demonstrated highly accurate and reliable results. The best performance of the model with the training accuracy of 98% and the testing accuracy of 85% was achieved using a balanced sampling method, fourteen predictor variables, the penalty parameter equal to 1, the radial basis function (RBF) kernel, and the value of 2 for the kernel's parameter. The urban expansion model based on SVM method can substantially improve the prediction accuracy and would be helpful for making appropriate plans and policies to mitigate the adverse impacts of urban expansion. Highlights: The capabilities of support vector machine (SVM) technique are explored to develop an urban expansion model. Three sampling methods are investigated to create an appropriate sample training dataset. A comprehensive combination of the most significant predictor variables is determined. Various SVM configurations are evaluated by regulating the 'penalty parameter, kernel function, and kernel's parameter. Novel goodness-of-fit metrics are proposed to specifically evaluate the performance of SVM model for LULC change modeling.