J. Intell. Syst. 20 (2011), 261 – 278 DOI 10.1515 / JISYS.2011.014 © de Gruyter 2011 Site Characterization Model Using Support Vector Machine and Ordinary Kriging Pijush Samui and Sarat Das Abstract. In the present study, ordinary kriging and support vector machine (SVM) have been used to develop three dimensional site characterization model of an alluvial site based on standard penetration test (SPT) results. The SVM is a novel type of learning machine based on statistical learning theory, uses regression technique by introducing "-insensitive loss function has been adopted. The knowledge of the semivariogram of the SPT values (N ) is used in the ordinary kriging method to predict the N values at any point in the subsurface of the site where field measurements are not available. The comparison be- tween the SVM and ordinary kriging model demonstrates that the SVM model is superior to ordinary kriging model in predicting N values in the site. Keywords. Ordinary kriging, semivariogram, support vector machine, standard penetra- tion test, site characterization. 2010 Mathematics Subject Classification. 60-XX. 1 Introduction Site characterization is an important task in geotechnical engineering practice and generally defined as the identification and description of the subsurface strata within the areas of influence of a project. The basic objective of site character- ization is to provide sufficient and reliable information and data on the site con- dition to a level of compatible and consistent with the needs and requirements of the project. It is necessary to predict geotechnical properties at any point of a site based on a limited number of tests. Prediction of geotechnical properties of a site is a difficult task for uncertainty. Uncertainty comes from spatial variabil- ity, measurement noise, measurement and model bias, and statistical error due to limited measurement [1]. Case studies have shown that predictions solely based on engineering judgment fails in 70% of the cases. As such, need for characteri- zation of soil strata based on more rational and scientific approach has been felt. The methods like wavelet based variance method, spectral density function, fractal model [2], Barlett’s statistics method [3], kriging [1,4–6] have been discussed to characterize soil stratification.