Enhancing a GIS Cellular Automata model of land use change using Support Vector Machine

Cellular Automata (CA) are dynamic mathematical systems (which are discrete in time and space, operate in uniform regular lattice, and are characterized by local interaction) that can be used with GIS to simulate land use change. Common cellular automata calibration is based on linear logistic regression. Linear models, however, are susceptible to over-fitting; especially when applied to nonlinearly separable data. This study presents the Support Vector Machine (SVM) for defining nonlinear cellular automata transition rules. SVMs avoid over-fitting by choosing a specific hyperplane among the many that can separate the data in the feature space. Historical Remote Sensing (RS) images serve as training data. The synergy between the GIS and CA is explored using the loose coupling technique. The model is applied to simulate land use change in Lekki area of Lagos, Nigeria from 1984 to 2000.