Traditional U-slot microstrip patch antenna designs heavily rely on iterative simulations based on expert knowledge, which results in high design costs, long cycles, and low design efficiency. To address these issues, this study proposes an innovative method that integrates gated recurrent unit (GRU) models with attention mechanisms to rapidly design U-slot microstrip patch antennas, enhance their local features, and improve design accuracy. Experimental results show that this method can not only quickly and accurately predict the antenna performance spectrum in the forward direction but also inversely deduce the antenna structure design based on the desired spectrum, demonstrating high accuracy in both forward and reverse predictions. This study not only provides a new perspective for the design of U-slot microstrip patch antennas but also offers certain reference value for the design of other antennas and communication devices.