Fisher discriminant and automatically getting transition rule of CA

This paper has put forward a new method for automatically getting transition rule of geographic Cellular Automaton (CA), basing on the combination of Fisher discriminant and discrete selection model. The core of CA is how to define conversion rule. However, at present heuristic methods are mainly adopted to define transition rule, greatly influenced by subjective factors. Having combined discrete selection model and improved Fisher discrimination, this model can successfully search the best variable combination to separate developing and non-developing units, and automatically give the parameter value of model. The results of comparison with Logistic regression model indicate that, this method has higher precision and the transition rule has clear physical meaning. In addition, this model has predominance in simulating multi-class complex change of land-use.