A comprehensive study is presented on leveraging machine learning to design and optimize an ultra-wideband (UWB) antenna. First, proposed a novel UWB antenna design using innovative geometries to achieve enhanced bandwidth and performance characteristics. Next, develop a data-driven approach, utilizing machine learning algorithms, to optimize the antenna's performance parameters, such as return loss. Through this approach, it is possible to overcome the limitations of traditional manual optimization methods and achieve highly efficient and customized designs for specific applications. In this work, machine learning techniques, including decision trees, random forest, gradient boosting, and XGBoost used and demonstrated the potential of ML-driven optimization to efficiently explore the design space and achieve superior results. The research highlights machine learning's transformative impact on antenna design, unlocking new possibilities and applications. The proposed antenna exhibits ultra-wideband characteristics, operating from 2.9 GHz to 16.37 GHz. Notably, it demonstrates five distinct bands with return loss values of -14 dB at 3 GHz, -28.87 dB at 4.3 GHz, -31.2 dB at 7.4 GHz, -18.07 dB at 9.14 GHz, and -29.76 dB at 14.2 GHz.