Aiming at the VMC1165B vertical machining center, the thermal characteristics and temperature field of the machine tool were analyzed, and based on experimental data,the thermal mechanism analysis and temperature field boundary condition calculation were avoided. Four robust temperature-sensitive points were selected by fuzzy clustering and Pearson correlation coefficient method to establish the thermal error model, and the predictive performance of the model was verified. It was compared with the thermal error prediction model of non-robust temperature-sensitive points selected by fuzzy clustering combined with grey correlation degree. The results show that the maximum residual of machine tool X axis of the robust temperature-sensitive points thermal error prediction model is reduced by 25.44%, while the maximum residual, mean absolute error and mean square error of the machine tool Y axis decrease by 25%, 23.03% and 33.25%, respectively.