The process of quality is traditionally considered non-productive. That is why the manufacturing industries aim to decrease inspection times to a bare minimum without sacrificing quality. With recent developments in Industry 4.0 technologies, data-driven in-situ quality control is a factor industry is increasingly relying on to minimize inspection times. Surface roughness parameter prediction has been the subject of a large body of research, but studies on the spatial surface roughness profile prediction are limited. This research contributes to this field by using vibration signals to predict the surface roughness profile pattern in-situ. A triaxial accelerometer mounted on the CNC spindle is used to capture the vibrations during a face milling process using a face milling tool with five indexable inserts. For one tool revolution, the observed acceleration in three axes and the surface roughness profile is periodic. Data-driven models are constructed to establish a correlation between the input signals and spatial surface roughness profile, i.e., by utilizing the machine learning models Gaussian Process Regression (GPR) and Bagged Ensemble Tree (BET). The results show a good correlation between the spatial surface roughness and the accelerometer signals. Furthermore, the addition of denoised velocities and displacements derived by the numerical postprocessing of the acceleration signals improves the performance of the models with minimum overfitting. With triaxial acceleration, velocity, and displacement as input data, the BET was the best performing model suitable for real-time monitoring of surface roughness with high accuracy and reliability.