This research investigates the optimization of Dielectric Resonator Antennas (DRA) using advanced machine learning (ML) techniques. The study addresses the inherent challenges in designing efficient DRAs, focusing on gain, bandwidth, and radiation efficiency parameters. This research proposes a new framework to automate the optimization process by integrating ML algorithms, significantly enhancing design efficiency and performance metrics. In the analysis, Neural Networks and k-Nearest Neighbors (k-NN) were utilized to predict and optimize key parameters of the DRA. Neural Networks demonstrated superior accuracy in predicting the S11 reflection coefficient, achieving over 95% accuracy, while k-NN provided quick and interpretable results, making it beneficial for initial parameter estimations. The combination of these ML techniques enabled efficient exploration and optimization of the DRA. Experimental validation was conducted through simulations using CST software, followed by performance analysis of the proposed antennas. The results indicated that the ML-optimized DRA achieved accurate prediction in S11 bandwidth.