Abstract: Least square vector machines had the quickly solving speed and the excellent characteristics to describe the little samples, nonlinear and high dimensions problem. But the regression accuracy and generalization performance of this method depend on a proper setting of its parameters. So, an optimal selection approach of LSSVM was presented based on particle swarm optimization (PSO) algorithm. Then , the prediction on the thickness of the loosen zone around roadway is made by PSO-LSSVM model, and the results show that this model has the characteristic of quickly convergence speed and highly calculating precision .So, this model is reasonable and feasible.

Key words: particle swarm optimization algorithm, least square support vector machines, thickness of loosen zone, prediction