As a key factor restricting the machining accuracy of computer numerical control (CNC) machine tools, thermal error is especially obvious in heavy CNC machine tools. Taking heavyduty floor boring and milling machine as an example, the temperature field characteristics of heavy-duty CNC machine tools are analyzed according to the thermal error measurement data.The temperature measurement point system was optimized based on the system clustering criterion considering both the correlation coefficient and Euclidean distance to reduce the collinearity between temperature measurement points. According to the optimized temperature measurement point, the thermal error prediction model of heavy CNC machine tool was established by multiple linear regression analysis. The field test shows that the thermal error prediction model can control the root mean square error within 10 micron and improve the thermal error prediction accuracy effectively.