ABSTRACT
Modeling pore-water pressure (PWP) responses to rainfall is an important part of monitoring hydrological behavior of hill slope. Of recent, soft computing techniques had been used to model these responses. Using support vector regression these responses can be modeled with very good accuracy. However, selection of appropriate kernel for such modeling is a necessity. The radial basis function was found to be the most suitable for modeling PWP responses, due to its competitive results and less complexity in implementation. In the mapping to higher dimensions, support vector machines can use any of several kernel functions. Kernel functions provide a simple avenue, which leads from non-linearity to linearity for algorithms, by simple expression of the dot product of the input. This makes it a good choice. Although the linear kernel Model has even fewer parameters. It lacks the capability to make good predictions and instead is relegated to predictions that constantly retain lagged records (one lag).
