Abstract:The accurate forecasting of urban taxi demands, which is a hot topic in intelligent transportation research, is challenging due to the complicated spatial-temporal dependencies, the dynamic nature, and the uncertainty of traffic. To make full use of the global and local correlations between traffic flows on road sections, this paper presents a deep learning model based on a graph convolutional network, long short-term memory (LSTM), and multitask learning. First, an undirected graph model was formed by considering the spatial pattern distribution of taxi trips on road networks. Then, LSTMs were used to extract the temporal features of traffic flows. Finally, the model was trained using a multitask learning strategy to improve the model’s generalizability. In the experiments, the efficiency and accuracy were verified with real-world taxi trajectory data. The experimental results showed that the model could effectively forecast the short-term taxi demands on the traffic network level and outperform state-of-the-art traffic prediction methods.

What problem does this paper attempt to address?

The paper aims to address the problem of urban taxi demand prediction, particularly focusing on the effective short-term prediction of taxi demand at the traffic network level. Specifically, the paper focuses on the following aspects: 1. **Problem Background**: In the research of Intelligent Transportation Systems (ITS), taxi demand prediction is a hot topic. Due to the uncertainty, dynamism, and complex spatiotemporal dependencies of taxi demand, accurately predicting when and where passengers will need a taxi becomes extremely challenging. 2. **Objective**: By analyzing and predicting the demand for taxi passengers, it is possible to more accurately reflect the dynamic changes in travel demand in different areas of the city. This not only helps passengers choose travel routes that save time and cost but also assists city managers in making informed traffic planning and public vehicle resource allocation decisions, thereby alleviating traffic congestion and reducing the waste of public resources. 3. **Technical Means**: To achieve this goal, the authors propose a deep learning model based on Graph Convolutional Network (GCN), Long Short-Term Memory (LSTM), and multi-task learning strategies. This approach can fully exploit the global and local correlations in traffic flow. 4. **Specific Methods**: - First, an undirected graph model is constructed to represent the spatial pattern distribution of taxi trips on the road network. - Then, LSTM is used to extract the temporal features of the traffic flow. - Finally, a multi-task learning strategy is employed to train the model to improve its generalization ability. 5. **Contributions**: - A new GCN-based graph network is proposed for predicting taxi departure and arrival flows. - Two types of traffic flow graphs are constructed, namely direct and indirect graphs, considering local and global correlations respectively. - Two related tasks—departure flow prediction and arrival flow prediction—are combined, and a multi-task learning strategy is adopted to enhance the learning process, avoid overfitting, and obtain more general results. - Comprehensive experiments are conducted using real-world taxi trajectory data, and the results show that the proposed model outperforms existing traffic prediction methods.

Multitask Learning and GCN-Based Taxi Demand Prediction for a Traffic Road Network

A Spatial-Temporal Gated Hypergraph Convolution Network for Traffic Prediction

Prediction of Urban Taxi Travel Demand by Using Hybrid Dynamic Graph Convolutional Network Model

Enhancing Demand Prediction: A Multi-Task Learning Approach for Taxis and TNCs

Dynamic Graph Convolutional Network-Based Prediction of the Urban Grid-Level Taxi Demand–Supply Imbalance Using GPS Trajectories

Convolutional Long Short-Term Memory Two-Dimensional Bidirectional Graph Convolutional Network for Taxi Demand Prediction

Deep Multi-View Spatial-Temporal Network for Taxi Demand Prediction

Region-Level Traffic Prediction Based on Temporal Multi-Spatial Dependence Graph Convolutional Network from GPS Data

Intelligent Prediction of Urban Road Network Carrying Capacity and Traffic Flow Based on Deep Learning

Application of an Adaptive Adjacency Matrix-Based Graph Convolutional Neural Network in Taxi Demand Forecasting

Graph learning-based spatial-temporal graph convolutional neural networks for traffic forecasting

T-GCN: A Temporal Graph ConvolutionalNetwork for Traffic Prediction

T-GCN: A Temporal Graph Convolutional Network for Traffic Prediction

Spatial–temporal short-term traffic flow prediction model based on dynamical-learning graph convolution mechanism

A spatial-temporal short-term traffic flow prediction model based on dynamical-learning graph convolution mechanism

Traffic Flow Driven Spatio-Temporal Graph Convolutional Network for Ride-Hailing Demand Forecasting

Passenger Demand Forecasting with Multi-Task Convolutional Recurrent Neural Networks.

Predicting taxi demands via an attention-based convolutional recurrent neural network

Taxi Demand Prediction Using Parallel Multi-Task Learning Model

Temporal Multi-Graph Convolutional Network for Traffic Flow Prediction

Multitask Hypergraph Convolutional Networks: A Heterogeneous Traffic Prediction Framework