Abstract: The support vector machine (SVM), as a statistical learning algorithm to minimize structure risk, is used to develop a forecasting model of drillability, thus avoiding the high-cost testing for determining plutonic rock stratum drillability and the overlearning which is often found in the application of neural network model of logging data. Learning the samples with SVM, the complex nonlinear mapping between rock stratum drillability and many logging data is given. To select the SVM parameters, the difference evolution (DE) algorithm as a global optimization algorithm is introduced to develop the DE-SVM evolutionary model so as to improve further the forecast precision by modeling. The numerical example indicates that the DE algorithm converges rapidly and its forecasting precision is higher than other conventional methods. The method proposed is of importance to selecting drill-bits for new oilwell and determining the drilling speed.