Cellular automaton urban growth simulating method based on random forest

Abstract

The invention discloses a cellular automaton urban growth simulating method based on the random forest. According to the method, based on the random forest algorithm, in the generation process of a decision-making tree, random factors are introduced into candidate space variables produced when sample sets and split nodes are trained; a transformational rule of an urban growth cellular automaton model is extracted and can be used for simulation and prediction of urban growth. The method has the advantages that prediction accuracy is improved on the premise that the operation amount is not obviously increased; the method is not sensitive to multicollinearity, the over-fitting phenomenon does not easily occur, and the method is well tolerant of the random factors in urban growth; error estimation outside a bag can be conducted, and model parameters can be rapidly obtained; space variable importance can be measured, and the effect of each space variable in urban growth is explained.

Description

technical field [0001] The invention relates to the field of geographic information, in particular to a cellular automaton city expansion simulation method. Background technique [0002] Cellular Automata (Cellular Automata, CA) was proposed by mathematicians Stanislaw M.Ulam and von Neumann in 1948. It was originally used to simulate the unique self-replication phenomenon of living systems. It is a simplified description of complex phenomena in nature. mathematical model. Since Tobler first applied cellular automata to urban expansion simulation in 1970, many scholars have carried out the application of urban cellular automata, and achieved rich results in many fields such as planning and ecology. [0003] The core of the cellular automata model is to define the transformation rules of cells. In each loop iteration operation, the transition rule is a dynamic function that determines the state of the cell at the next moment by the current state of the cell and its neighbor...

AI Technical Summary

Problems solved by technology

Neural network can better solve nonlinear classification problems, but this method belongs to the black box structure, it is difficult to understand the physical mechanism, and there is a problem of slow convergence

Ant colony intelligence algorithm has strong self-adaptive ability, but it has defects such as high computational cost, slow convergence speed, easy to fall into local optimum, etc.

Images