Prediction of Taxi Demand Based on ConvLSTM Neural Network

As an important part of the urban public transport system, taxi has been the essential transport option for city residents. The research on the prediction and analysis of taxi demand based on the taxi GPS data is one of the hot topics in transport recently, which is of great importance to increase the incomes of taxi drivers, reduce the time and distances of vacant driving and improve the quality of taxi operation and management. In this paper, we aim to predict the taxi demand based on the ConvLSTM network, which is able to deal with the spatial structural information effectively by the convolutional operation inside the LSTM cell. We also use the LSTM network in our experiment to implement the same prediction task. Then we compare the prediction performances of these two models. The results show that the ConvLSTM network outperforms LSTM network in predicting the taxi demand. Due to the ability of handling spatial information more accurately, the ConvLSTM can be used in many spatio-temporal sequence forecasting problems.