Abstract: The thermal stability of CNC machine tools is a core factor restricting their machining accuracy and long-term reliability. Taking a precision horizontal milling machining center as the research object, the heat generation characteristics of key machine tool components (such as spindle, feed axis motor, and ball screw) and the evolution law of thermal errors were analyzed using simulation software. The distribution of core heat sources, the time sequence for the machine tool to reach thermal equilibrium, and the period of rapid thermal error fluctuation were identified. Combined with magnetic adsorption K-type thermocouples (covering key temperature measurement points) and laser measurement equipment, the measured temperature-thermal deformation data of heat-prone components are collected to verify the accuracy of the simulation conclusions. On this basis, suitable models are constructed for different types of thermal errors: multiple linear regression is used for modeling the linear thermal errors of the feed axis, and support vector regression is adopted for modeling the nonlinear thermal drift of the spindle. Both types of models have high prediction accuracy, and the support vector regression model performs better in nonlinear error prediction. This research provides an efficient thermal error prediction scheme for machine tool manufacturing enterprises, clarifies the key targets and key periods for thermal management, solves the problem of insufficient accuracy of traditional single models, and provides theoretical support and experimental basis for improving the thermal stability, thermal robustness, and machining accuracy of precision horizontal milling machining centers.