基于LSTM神经网络融合用户相似性的移动轨迹预测

With the popularity of smart devices and location acquisition technology, a large amount of trajectory data has been accumulated. By learning the trajectories of people, we can get the information about their movement patterns. Based on this information, trajectory prediction is very valuable for some problems such as traffic planning and urban planning. In this paper, we propose a users\' trajectories prediction model by using LSTM neural networksbased on user similarity. By analyzing users\' moving trajectories, firstly we mine users\' movement patterns and calculate their similarities. Next, users are divided into communities based on similarities. Then, LSTM trajectory prediction networks of the community users and the individual user are constructed and the prediction results of these two networks are weighted and summed up to obtain the prediction of the user\'s next destination point. In this paper, the LSTM neural network is used to model sequences and the information provided by both the history of individual moving trajectories and user similarity is taken into account. The experimental results indicate that the prediction results have high accuracy.