The general literature in SVM has not discussed in detail the subject of tuning the various user-defined parameter in SVMs. In this chapter the theoretical implication of these parameters is discussed. Experimental results are included to confirm the expected behaviour of SVM when these parameters are changed. SVM has only two major parameters which are defined by the user. There is the trade-off between the margin width and the classification error (C in Eqn. ), and the kernel function. Most kernel functions will also have a set of parameters.