This research aims to identify the boundary models of misfire and knocking by applying the Support Vector Machine (SVM). It is necessary to select the kernel functions appropriately in utilizing the SVM. The linear kernel, the Radial Basis Function kernel, the polynomial kernel and the sigmoid kernel are utilized for the kernel functions of the SVM. The boundary models of misfire and knocking are identified by using the half of training data to reduce the number of measurement points. Moreover, an input is represented by other inputs to identify accurate regions from less measurement points. An effectiveness of the kernel functions is verified by calculating the accuracy of the identified boundary models.