SASRM: A Semantic and Attention Spatio-temporal Recurrent Model for Next Location Prediction

Abstract

Predicting user's next location is of great importance for a wide spectrum of location-based applications. However, most prediction methods do not take advantage of the rich semantic information contained in trajectory data. Meanwhile, the traditional LSTM-based model can not capture the spatio-temporal dependencies well. In this paper, we propose a Semantic and Attention Spatio-temporal Recurrent Model (SASRM) for next location prediction. Firstly, the SASRM put forward a method for encoding semantic vectors and concatenating vectors (location, time and semantic vectors) as input to the model. To capture the spatio-temporal dependencies, we design a variant recurrent unit based on LSTM. Further, an attention layer is used to weight hidden state to capture the influence of the historical locations on the next location prediction. We perform experiments on two real-life semantic trajectory datasets, and evaluation results demonstrate that our model outperforms several state-of-the-art models in accuracy.