A Multi-Task Matrix Factorized Graph Neural Network for Co-Prediction of Zone-Based and OD-Based Ride-Hailing Demand

Ride-hailing service has witnessed a dramatic growth over the past decade but meanwhile raised various challenging issues, one of which is how to provide timely and accurate short-term prediction supply demand. While predictions for zone-based demand have been extensively studied, much less efforts paid origin-destination (OD) based (namely, originating from zone another). However, OD-based even more important worth further explorations, since it provides elaborate trip information in near future as reference fine-grained operations, such routing matching shared ride-hailing services that pick up drop off two or passengers each ride. Simultaneous both can be an interesting practical problem platforms. To address issue, we propose multi-task matrix factorized graph neural network (MT-MF-GCN), consists major components: (1) GCN (graph convolutional network) basic module captures spatial correlations among zones via mixture-model (MGC) network, (2) factorization By evaluations on real-world on-demand data Manhattan Haikou, show proposed model outperforms state-of-the-art baseline methods zone- predictions.