<p>Device to Device (D2D) communication is a viable approach for enhancing performance as it improves communications with or without backbone infrastructure. Device discovery (DD) is an essential process in D2D communication that makes it possible for devices to find and connect with one another. While number of algorithms and methods for DD has been researched, proposed and published, this research is novel attempt to incorporate machine learning for DD in device to device communication. In the present research, we propose a powerful machine learning based model and technique employing Feed Forward Neural Network and Decision Tree for D2D discovery to predict the location of devices. Feed Forward Neural Network has been implemented with MLP regressor. The model is initially implemented with 3 hidden layers, and later further tuned to 5 hidden layers to reduce mean square error to 0.94e-3. The decision tree algorithm employs training with 5 and 10 hidden layers reducing mean square error to 9.26e-6.The paper evaluates the proposed algorithm using simulations and results show that the machine learning algorithms converge to predict the DD in near real time thus improving the efficiency and accuracy of D2D discovery. The proposed approach demonstrates promising performance in terms of classification accuracy and scalability, showcasing its potential for practical implementation in real world D2D communication scenarios. The proposed algorithm can be useful for various applications that require efficient and accurate DD.</p>