中文摘要：

针对目前广泛应用的灰色理论、遗传算法（GA）和人工神经网络（ANN）等方法预测隧道稳定性的缺陷，提出应用稳健性能较好的ε-SVR（support vector machine）算法对非对称连拱隧道围岩位移演化规律进行预测研究。应用加速混合遗传算法搜索ε-SVR最优参数，以提高ε-SVR的预测能力。将预测结果与灰色理论、BP神经网络预测结果进行比较，显示ε-SVR算法学习和预测精度高。

英文摘要：

Because the prediction accuracy of gray theory, GA and ANN algorithm is insufficiency for tunnel rock surrounding stability, the method of ε-support vector machines was applied to researching of evolution law for tunnel rock surrounding displacement; and in order to enhance the learning efficiency of ε- support vector machines and the capability of forecasting, the accelerated hybrid genetic algorithm （GA） is used for optimizing parameters of ε-support vector machines. Comparison the forecasting results of gray theory, GA and ANN and monitoring results for tunnel rock surrounding displacement, the results show the learning efficiency and prediction accuracy of ε-support vector machines is superior to gray theory, GA and ANN obviously.