Predicting origin-destination ride-sourcing demand with a spatio-temporal encoder-decoder residual multi-graph convolutional network

Real-time Forecasting of Dockless Scooter-Sharing Demand: A Context-Aware Spatio-Temporal Multi-Graph Convolutional Network Approach

“…The deep learning methods have provided researchers powerful tools to deal with travel demand prediction problems, such as taxi demand prediction (Liu et al, 2019;Xu et al, 2017;Yao et al, 2018), ride-hailing demand prediction (Geng et al, 2019), ridesourcing demand prediction (Ke et al, 2021), and bike-sharing demand prediction (Kim et al, 2019;Lin et al, 2018). Since the demand varies spatially and temporally, different deep learning methods were used to capture spatial dependency and temporal dependency in these studies, and the results showed that these deep learning methods outperformed the classical machine learning models (e.g., RF and GBDT) and the statistical models (e.g., linear regression and ARIMA).…”

“…Since the factors discussed above can greatly influence the usage of dockless scootersharing, these factors should be considered when forecasting the dockless scooter-sharing demand. Although some existing demand forecasting models has taken some of these factors into account (Ke et al, 2021;Tang et al, 2021), few studies comprehensively included all kinds of these factors in their models when predicting the real-time dockless scooter-sharing demand, especially the weather conditions.…”

Recurrent origin–destination network for exploration of human periodic collective dynamics

“…Toqué, Côme, El Mahrsi, and Oukhellou (2016) were among the first to use the long short-term memory (LSTM) neural network model to estimate future OD flows by using the historic OD flows as input. Since their work, more complex deep learning models have been proposed to estimate dynamic future flows, among them convolutional LSTM (Duan et al, 2019), contextualized spatial-temporal networks (Liu et al, 2019), dual-stage graph convolutional recurrent neural networks (Hu, Yang, Guo, Jensen, & Xiong, 2020), spatial-temporal LSTM (Li et al, 2020), spatial-temporal encoder-decoder residual multi-graph convolutional networks (Ke et al, 2021), and dynamic node-edge attention networks (Zhang, Xiao, Shen, & Zhong, 2021). Therefore, dynamic OD networks provide a more comprehensive and detailed day-to-day description for urban dynamics and have become a powerful tool for understanding collective dynamics in recent years.…”

A Lightweight Approach for Origin-Destination Matrix Anonymization

“…In this paper, we consider trajectories defined only by their first and last spatiotemporal points, i.e., an Origin/Destination (OD)-matrix. Although they leave out most of the trajectory information, OD-matrices are a key element in the transport analysis framework as they can be used to understand the dynamic of transport demand [4], make long-term prediction [5] and allow for transport simulations [6]. Anonymization of OD-matrices has been explored in [7] with a uniform aggregation to achieve k-anonymization, and in [8,9] with a tree-based spatial aggregation to achieve differential privacy.…”