支持向量回归机在滑坡安全可靠度评价中的应用研究

Study on application of support vector machine in slope realibility analtsis

探讨支持向量机的参数优化方法,将学习能力较强的支持向量机引入滑坡可靠度评价当中,并根据数理统计的理论,对滑坡稳定性敏感性因素作数据处理,并证明其合理性,在此基础上提出一种改进的可靠度指标的计算方法。为便于对比,分别选取49个学习样本和121个学习样本两种计算模型,计算实例表明,这两种计算模型与蒙特卡洛模型计算结果非常接近,充分体现了在小样本的情形下支持向量机的优势,说明将其应用于滑坡安全可靠度评价中是可行的。

YANG Dong,JIANG Li （ Key Laborato~＇ of Water Conservancy and Water Resources in Anhui Province, Bengbu 233000, Anhui Huaiwei Sci- ence Institute of Water Conservancy in Anhui Bengbu 233000, Anhui） Absrtact： The optimization of the parameters of SVM is discussed in this article. The support vector machine which has the strong learning ability is applied to evaluate the stability of landslides. In accordance with the theory of mathematical statis- tics, the data of the sensitivity factors of landslide stability are processed, which is proved to be reasonable. And an improved method of calculating reliability index is proposed here. We choose two different learning computing models, one including 49 samples, tile other 121 samples . Calculation examples show that the calculation of these twn models are very close to the results of the Monte - Carlo model , which reflects tire advantages of support vector machine when the number of samples is insuffi- cient, and that it is feasible to be applied to evaluate the reliability of the landslide safety.