Relevance Vector Machine for Prediction of Soil Properties

Abstract

One of the first, most important steps in geotechnical engineering is site characterization. The ultimate goal of site characterization is to predict the in-situ soil properties at any half-space point for a site based on limited number of tests and data. In the present study, relevance vector machine (RVM) has been used to develop three dimensional site characterization model of an alluvial site based on standard penetration test (SPT) results. In three dimensional analysis, the function N=N(X,Y,Z) where X, Y and Z are the coordinates of a point corresponds to SPT value(N), is to be approximated with which N value at any half space point in site can be determined. RVM provides an empirical Bayes treatment of function approximation by kernel basis expansion.  It has also capability to estimate the prediction variance. The potential of RVM for prediction of N value is assessed in this study.