Cultivated land change forecast based on genetic algorithm and least squares support vector machines

A prediction method of cultivated land change based on least squares support vector machines (LS-SVM) was developed by studying the inherent tendency toward land change and simulating the trajectories of changes in land use. A nonlinear dynamic model of cultivated land change and influence factors was introduced. The prediction accuracy was improved by using the genetic algorithm to automatically determine the optimal parameters of least squares support vector machines. The proposed model has been thoroughly tested on predicting the cultivated land change during the period of 1987－2000 in Wuxi, Jiangsu. The results were compared and analyzed with those obtained from multiple regression, GM(1,1), BP algorithm, support vector machines(SVM) and the survey data on cultivated land change. The evaluation of prediction precision showed that the method based on LS-SVM was far more accurate than multiple regression, GM(1,1) and BP network model. Compared with the support vector machines model, the method was even slightly better and possesses much less algorithm complexity and higher computational efficiency. The overall performance suggests that the method is effective in predicting land change.