RFID technology has gained significant attraction across industries by enabling efficient inventory management and accurate asset tracking. Antennas play a crucial role in facilitating seamless data exchange in RFID tags. Different antenna types have been widely employed in RFID tags including monopole, dipole, spiral, helical, patch and loop antennas. Among these, loop antenna is notable for its compact size and suitability for various orientations, making it ideal for applications needing flexible antennas. To make the antenna implantable in RFID tags the size of the antenna have to be reduced without compromising the bandwidth and Radio Frequency (RF) performance. In this paper, a gradual transition in circular loop is given, by printing an encapsulation at the centre of the antenna is proposed to improve the antenna bandwidth, allowing for effective communication in the frequency range of 3.5–9.25 GHz. Conventional antenna design and optimization methods can become laborious and difficult due to the complexities of antenna designs. To overcome this challenge, this paper also proposes the utilization of Machine Learning regression model, particularly XG Boost, which has demonstrated remarkable accuracy in predicting antenna performance, based on physical parameters. Therefore, it substantially expedites the optimization process compared to conventional Electro Magnetic (EM) simulators. The proposed XG Boost approach predicts the antenna performance in 10 s and reduces the design time of the antenna by 99.16% compared to the conventional EM Simulator which takes approximately 20 min for each simulation.