In this work, the design and realization of a circular reflectarray antenna (CERA) with semi- elliptic rings are carried out for X-band. In order to achieve a fast design optimization process, 3D EM simulated data-driven multilayer perceptron neural network (MLP NN) model of the unit element is generated providing the unit reflection phase as a continuous function of the radii and gap of the ring for the width w = 0.5 mm on the FR4 with SMALL ELEMENT OFr = 4.4, h = 1.5 mm. The proposed CERA consists of 148 unit cells taken symmetrical both in the Y and X-axis. Thus, a meta-heuristic optimization algorithm honey bee mating optimization (HBMO) is applied to 21 elements between 10 and 12 GHz to determine their optimal geometries for the required reflection angles according to the phasing scheme. The simulated and measured performances of the designed CERA have been found in good agreement and its performance is compared also with its counterpart designs in literature and resulted to exhibit superior performance.