In this paper, a novel technique is proposed for localization of Subthalamic Nucleus (STN) during deep brain stimulation (DBS) Surgery. DBS surgery is performed on individuals living with Parkinson's disease (PD) to permanently implant stimulation electrodes for managing some motor symptoms of PD. The most challenging part of this surgery is to accurately place the electrodes inside the STN. Commonly, microelectrode recordings (MERs) are interpreted by the surgical team intraoperatively to estimate the location of electrodes and detect the borders of the STN. In this work, we aim to automate the process of localizing the STN using a machine learning technique (trained based on the electrophysiological signals that we have collected during 20 surgeries). The proposed approach is capable of detecting the dorsal borders of the STN during the procedure with high accuracy (85%), and outperforms the current state-of-the-art approach for this application.