Neighborhood-Augmented LSTM

In this paper, we propose an approach for predicting the pick-demand of a given taxi-stand, that takes into account not only the demand-history of the particular stand but it also considers information from neighboring stands.

Our model is an LSTM neural network augmented with information from the spatial neighborhood of the stands. Experiments with two versions of the taxi demand dataset from the city of Porto, Portugal show that our approach can provide better predictions comparing to approaches that do not exploit the neighborhood.

Tai Le Quy, Wolfgang Nejdl, Myra Spiliopoulou, Eirini Ntoutsi. A Neighborhood-Augmented LSTM Model for Taxi-Passenger Demand Prediction. In International Workshop on Multiple-Aspect Analysis of Semantic Trajectories - MASTER 2019 @ ECML-PKDD 2019..