Abstract—Recently, unmanned aerial vehicles (UAVs) have found numerous telecommunication applications due to their high feasibility and low cost. Optimizing the UAV communications system requires determining the characteristics and sensitivity of wireless signals to propagation effects in different environments, and frequency bands. Hence, accurate path loss prediction models are vital for planning, evaluating, and optimizing UAV-based communication networks. This research proposes a path loss prediction model for UAV-to-UAV channels using two variants of the LSTM deep learning algorithm: bidirectional long short-term memory (LSTM) and encoder-decoder LSTM with hyperparameter tuning. The proposed model has been assessed using metrics such as mean absolute error (MAE), root mean square error (RMSE), and R-squared (R2). The proposed model has higher accuracy when compared with a traditional empirical model, and earlier machine learning models.