RNN-Based User Trajectory Prediction Using a Preprocessed Dataset

Future mobile networks are rightly expected to face the prospect of limited available resources. Continuous technological advances and growing number of mobile devices highlight the importance of further improving the performance of mobile networks. User mobility poses technical problems in network management. It is essential to ensure a satisfactory level of quality of service for users. To achieve this goal, self organizing networks (SONs) are potential solutions to fulfill the requirements of users using learning algorithms. In this paper, we propose an intelligent mobility model to predict future trajectory of the mobile user in mobile networks. The proposed approach has two main parts, including mobility data preparation and user mobility prediction. Our primary focus is on providing a carefully tailored mobility data from raw mobility datasets using line simplification techniques. Next, we use the accurately prepared data for learning user mobility behaviour and predicting user future trajectory using recurrent neural networks and its variants. Simulation results show a substantial decrease in execution time from 4616s to 932s for the best case. The proposed learning approach obtains a loss value of 0.10 using a model based on long short term memory (LSTM).