Predicting user trajectories using deep learning algorithms / Ahmad Zaki Aiman Abdul Rashid, Azita Laily Yusof and Norsuzila Ya’acob

In order to produce seamless handover performance, a user’s trajectory acts as a catalyst in determining the exact time and position of making the handover from one base station to another base station. Due to this, this paper predicts user’s future trajectory from past trajectory utilizing deep learning (DL) algorithms which are Long-Short Term Memory (LSTM), BiDirectional LSTM, and Gated Recurrent Unit (GRU). Next, the performance of the model will be evaluated using regression metrics such as Mean Squared Error (MSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and the Coefficient of Determination (R2). The simulation results displayed LSTM model surpasses other models (GRU, BiDirectional LSTM) on the basis of accuracy achieved such as lowest MSE (0.084), MAE (0.254), MAPE (83.6%) with the highest R2 score (-0.379). Our LSTM model was also compared to other researchers LSTM-based model for trajectory prediction and produce greater accuracy with ADE of 0.2359 and FDE of 0.1834. These conclude that LSTM model are the most suitable model for predicting user trajectories among DL algorithms. This work demonstrates the potential of the LSTM model for predicting user trajectories with high accuracy and improve handover performance through prediction.