This study presents a novel printed antenna design that operates at the millimeter-wave frequencies of 28 and 34 GHz, which are crucial for the current and upcoming mobile communication generations. The radiating component in the antenna is a slot-etched rectangular ring that is fed through a stepped impedance microstrip line feed. Using advanced machine learning techniques, the design parameters of the suggested antenna have been fine-tuned to ensure optimal impedance matching at 28 GHz within the frequency range of 27.61-28.49 GHz. Additionally, the antenna also provides excellent impedance matching at 34.5 GHz within the frequency range of 33.61-34.27 GHz. Using the designated antenna, a Multiple Input Multiple Output (MIMO) system with two ports is constructed. The MIMO system's performance is evaluated by analyzing channel capacity loss (CCL), diversity gain (DG), and envelope correlation coefficient (ECC), which showcases outstanding outcomes. The study further explores the optimization of a antenna's structure using a Taguchi-based Neural Network (Taguchi NN) approach to predict the reflection coefficient (|S