Abstract: Thermal error seriously affects the machining accuracy of the machine tool. The thermal characteristic analysis of the key parts of the machine tool is an important link in the development of precision machine tool. Therefore, this paper studies the relationship between temperature rise and positioning error by measuring the thermal characteristics including the temperature and positioning error of special position of CNC machine tool, and proposes a thermal error modeling method based on Bayesian neural network. Using K-means clustering and correlation coefficient method to select temperature sensitive points can effectively suppress multicollinearity between temperature measurement points. The results show that the accuracy of machine tool can be improved by 88.015 9% by using bayesian neural network, which is 15.763 8% higher than BP neural network. Compared with BP neural network model, bayesian neural network has better prediction performance. Bayesian neural network model provides a new idea to reduce the influence of thermal error of machine tool.