Thermal error caused by the thermal deformation is one of the most important concerns influencing the accuracy of machine tools. In this investigation, a thermal-structure model is developed by the finite element method (FEM) to effectively simulate thermal characteristics for predicting the transient temperature field and thermal deformation of a high speed spindle system. The FEM predictions were then compared against the measured temperatures and deformations at the tool center point (TCP) for model validation. Since the machining error must be estimated to achieve high quality and precision of machining products, this study applied the design of experiment (DOE) method to predict machining errors. The Pearson’s correlation analysis showed that the temperatures measured from the specific detection points were more significant than those from other points on the machining errors. The regression model was satisfactory to forecast the response variables formed by MATLAB. Alternatively, the significance of factors and the fitness of the designed model were verified by the ANOVA and residual analyses in conjunction with the collinearity diagnostics. The implementation of this model in the machine tool under study is expected to greatly reduce its thermally induced deformation up to 96.5% in composite working conditions for a spindle system.