<p>Thermal deformation is a critical factor that limits the accuracy and long-term stability of computer numerical control (CNC) grinding machines, particularly under high-precision, long-cycle operations. In this study, the effectiveness of an oil cooling system in mitigating spindle-induced thermal displacement and improving machining accuracy in a column-type CNC hydraulic surface grinding machine was evaluated. A multisensor measurement framework was established to record temperatures at ten key locations (T1–T10) and the corresponding three-axis displacements (X, Y, and Z) during prolonged operation under both cooling-off and cooling-on conditions. The collected data were analysed using statistical and data-driven methods, including principal component analysis (PCA) to characterize the global thermal distribution, Gaussian mixture models (GMMs) to classify thermal operating states, and random forest regression to predict thermal displacement with high accuracy (with R² values that exceeded 0.95). Feature importance analysis revealed the spindle motor seat (T5) and motor body (T7) to be the dominant thermal contributors to structural deformation. The experimental results show that oil cooling reduced the average steady-state thermal displacement by approximately 10.4% across all the axes. The normalized peak displacement errors decreased from 13.5%, 7.9%, and 15.2% to 0.9%, 0.2%, and 4.7%, respectively. Based on these findings, a regression-based thermal error compensation model that incorporates weighted temperature features was developed to capture the nonlinear coupling between thermal inputs and mechanical responses. The integrated framework provides a robust and industrially applicable solution for improving thermal stability and machining accuracy in high-precision CNC grinding applications.</p>