The present work proposes a data-driven method to address the issue of increased computational time in predicting the thermal field in the Powder Bed Fusion process (PBF). Understanding the temperature field evolution is essential in metal Additive Manufacturing (AM) process like PBF technology to produce the right quality parts. As the process of metal AM involves physical phenomena, including fluid flow, heat and mass transfer, as well as structural loads, the underlying physics is complex and difficult to predict temperature fields in reasonable time with increased accuracy. Physical experiments and numerical simulations can be expensive and time-consuming. Hence a data-driven approach has been introduced to address the issue of increased computational time to predict the thermal field developed during the process. The methodology proposed in this work consists of a thermal model and a surrogate model based on Gaussian Process Regression (GPR). Initially, the transient thermal behaviour is studied based on Finite Element Analysis (FEA). Later, a surrogate model based on the Gaussian process is developed from the FE simulated data to decrease the computational costs of high-fidelity physics-based simulations. The GPR model has predicted the thermal fields in less time than that of the physics-based FEA model. To validate our approach, both the numerical simulations and GPR-based model are justified by conducting single-track experimentations using Inconel 625 (IN625). The results have shown that the proposed GPR model predictions are in good agreement with the temperature measured from experiments. The studies have revealed the influence of process parameters like scan speed and laser power on the thermal field. Single track dimension for laser exposure seem to correlate with GPR predicted track width. This prediction could be employed to predict distortion, residual stress and other process response in a shorter time.