Acta Geographica Sinica ›› 2010, Vol. 65 ›› Issue (6): 656-664.doi: 10.11821/xb201006003

Previous Articles     Next Articles

A Nonlinear Polynomial Model for Urban Expansion Incorporating Genetic Algorithm and Support Vector Machines

ZHANG Hao1, LUO Yi-yong2, ZHANG Li-ting2   

  1. 1. College of Civil Engineering and Architecture, Zhejiang University of Technology, Hangzhou 310032, China;
    2. East China University of Technology Yangtze College, Nanchang 330013, China
  • Received:2010-01-17 Revised:2010-02-08 Online:2010-06-25 Published:2010-06-25
  • Supported by:

    National Natural Science Foundation of China, No.40874010; Open Foundation of Jiangxi Provincial Key Laboratory of Digital Land, DLLJ201014

Abstract:

With comparative analysis of strengths and weaknesses of current urban expansion simulation models and nonlinearly combining the advantages of best individual-based models, a nonlinear polynomial model of urban spatial expansion has been proposed by using the powerful functions of support vector machine to describe highly complicated nonlinear systems and then to make a better fit for them. The accuracy of the proposed model has been effectively improved by using the parameters of support vector machine optimized with genetic algorithm to reduce the negative influence, exerted by the non-rational design of parameters, on the modeling accuracy of support vector machine. With analyzing the relationship between the error arising from the combined model and all individual-based models, we conclude the ways to improve the accuracy of the nonlinear polynomial model of urban expansion equipped with support vector machine as follows: The first is to improve the accuracy of individual-based models; the second is to enlarge differences between individual models. In the case study of Changsha city, individual-based simulation models of urban spatial expansion constructed by multiple regression model, GM(1,8), BP network and LS-SVM are used to build a linear combination model of urban spatial expansion and a nonlinear combination model of urban spatial expansion equipped with genetic algorithm and support vector machine. A comparison of accuracy of selected models shows that the accuracy of nonlinear polynomial model of urban expansion equipped with genetic algorithm and support vector machine is much higher than any individual-based simulation model, and also higher than the linear combination model, and therefore, an efficient new model of urban expansion is established.

Key words: urban expansion, genetic algorithm, support vector machine, nonlinear combination, accuracy analysis, Changsha