Abstract: For the extremely complex relationships among influencing factors in geotechnical engineering, the support vector machine (SVM) was introduced in and the basic algorithm of support vector regression (SVR) was presented for complex geotechnical system modeling predicting in practical applications. The powerful capabilities of SVR for efficient self-learning over, small data sets and solving highly nonlinear problems were illustrated with an estimation of structural response to the mining-induced blast vibration of structures around a mine. The satisfactory results showed that the method proposed provides a feasible and promising alternative to deal with the complexities in geotechnical engineering. In addition, some experiments were performed to examine the sensitivity of parameter selection in SVR, with some important conclusion and advice obtained for further applications.