A Deep Learning Approach to Forecasting Short-Term Taxi Demands

In this study, we investigate the short-term taxi demand prediction. We develop a deep learning based approach that utilizes Long Short-Term Memory (LSTM) units in our artificial neural network. We compare our method against well-known time series prediction baselines in the literature. Using a dataset collected in real life, we show that our deep learning based method performs better than the baselines in forecasting short-term taxi demands in Manhattan, NYC.