Abstract—For intelligent wireless communication systems, precise wireless channel estimation and the development of a digital twin capable of replicating wireless channel characteristics are essential. Standard channel‑estimation methods such as Least Squares (LS) and Minimum Mean Squared Error (MMSE) are built upon linear estimation techniques and thus struggle to capture the nonlinear and temporal dynamics of wireless channels. This work focuses on capturing the temporal variations of the wireless channel by developing a digital‑twin framework based on a Long Short‑Term Memory (LSTM) neural network. The LSTM is trained with real‑world channel measurement data to learn complex relationships among wireless channel characteristics. Predicting channel characteristics helps reduce channel‑estimation overhead and improves the accuracy of channel‑state estimation. The proposed framework achieves a Mean Squared Error (MSE) of 0.035 and a correlation coefficient of 0.999, outperforming conventional channel‑estimation approaches including LS, MMSE, and the Gated Recurrent Unit (GRU)‑based model. The results demonstrate that the LSTM‑driven digital twin can accurately replicate the temporal and amplitude characteristics of the physical wireless channel.  
 
Keywords—digital twin, Long Short-Term Memory (LSTM), channel prediction, Mean Squared Error (MSE)