Origin-destination (OD) flow and demand prediction for metro and taxi systems

Origin-destination (OD) prediction forecasts the flow or demand between pairs of stations or regions, which supports real-time traffic management, vehicle dispatch, and resource allocation in intelligent transportation systems. Two papers (Yu et al., Expert Systems with Applications 2025; Gong et al., Expert Systems with Applications 2026) study this problem on metro and taxi data.

BiST-IF (Expert Systems with Applications 2025) addresses the delay of OD flows and uses arrival OD (Out-OD) flows as a supplementary signal, through OD delay correction, bi-directional attention, and mutual information flow evolution. DSTCN (Expert Systems with Applications 2026, open access) models dynamic, coupled spatio-temporal patterns with three modules, Glstm2D, Simformer, and FF-TM, and is evaluated on NYC-TOD2018, NYC-TOD2019, and HZMetro.

Papers compared

DSTCN: Exploiting dynamic spatio-temporal correlations for origin-destination demand prediction

Yongshun Gong, Piao Yu, Xu Zhang, Xinxin Zhang, Xiushan Nie, Haoliang Sun · Expert Systems with Applications, vol. 299, article 130095 (2026)

DSTCN (Dynamic Spatio-Temporal Correlation Network) forecasts origin-destination demand matrices with three modules: Glstm2D for origin- and destination-side demand trends, Simformer for Transformer-based spatial similarity across the OD matrix, and FF-TM for feature fusion and temporal modeling. On NYC-TOD2018 its MAE of 1.469 is 3.04% below the best baseline.

• Glstm2D pairs a destination-side branch (D-Glstm) with an origin-side branch (O-Glstm). Each combines a BiLSTM with spatio-temporal graph learning (STGL) to model demand trends along both axes of the OD matrix.
• Simformer merges each 4 x 4 block of OD cells, adds one-hot day-of-week and hour-of-day features, and applies multi-head self-attention. FF-TM fuses both feature streams, then models time with a GRU and multi-head temporal convolution.
• On NYC-TOD2018 and NYC-TOD2019 (69 Manhattan taxi zones, hourly), DSTCN reaches MAE/RMSE 1.469/3.284 and 1.273/2.909, 3.04% to 4.56% below the best of 12 baselines on each metric (p < .001).
• On HZMetro (80 Hangzhou metro stations, 10-minute intervals), DSTCN has the lowest MAE and RMSE at 10, 30 and 60 minutes, with MAE 0.931, 0.924 and 0.935, 9.61% to 11.79% below the best baseline.

BiST-IF: Enhancing origin–destination flow prediction via bi-directional spatio-temporal inference and interconnected feature evolution

Piao Yu, Xu Zhang, Yongshun Gong, Jian Zhang, Haoliang Sun, Junjie Zhang, Xinxin Zhang, Yilong Yin · Expert Systems with Applications, vol. 264, article 125679 (2025)

BiST-IF predicts origin–destination (OD) flows between metro stations or urban areas by correcting delayed recent OD matrices, applying bi-directional origin/destination attention, and fusing arrival-side (Out-OD) flows through an attention-based mutual information mechanism. It lowers MAE by an average of 7.55% on HZMetro relative to the best baseline.

• OD Delay Correction (ODDC) completes delayed recent OD matrices using period delay probability and real-time flow features. Without it, HZMetro RMSE rises from 1.799/1.797/1.797 to 1.804/1.809/1.827 (10/30/60 min).
• OD flow Bi-directional Attention (ODBA) combines BiLSTM trend extraction, attention between origin and destination features, and a spatio-temporal feature extraction layer, applied to weekly, daily and recent OD inputs (ratio 2:1:2).
• Mutual Information Flow Evolution (MFE) uses weekly Out-OD features as the key in 4-head attention over OD features; a new gated feature update module then stabilises the periodic pattern.
• On HZMetro and NYC-TOD2018, BiST-IF beats nine baselines on every metric and horizon, with average MAE/RMSE reductions of 7.55%/7.31% and 5.77%/2.33% against the best baseline.

Key concepts

Origin-destination (OD) prediction

Forecasting the matrix of trips from every origin station or zone to every destination in the next time interval. Papers: BiST-IF, DSTCN

OD delay (incomplete OD data)

At prediction time, trips still in progress have no known destination, so the latest OD matrices are incomplete. Papers: BiST-IF

HZMetro

Hangzhou metro passenger flows for 80 stations in January 2019. Papers: BiST-IF, DSTCN

NYC-TOD

New York City yellow-taxi OD demand between Manhattan zones (2018 and 2019 versions). Papers: BiST-IF, DSTCN

Frequently asked questions

Which paper addresses the delay of origin-destination flows?

BiST-IF (Expert Systems with Applications 2025). When recent OD data are collected, some passengers are still in transit, so the latest OD matrices are incomplete; BiST-IF corrects this OD flow information using period delay probability and real-time flow features, and models the interplay between OD and arrival (Out-OD) flows.

Which OD demand prediction paper is open access?

DSTCN, published in Expert Systems with Applications volume 299 (2026) under a CC BY licence, DOI 10.1016/j.eswa.2025.130095.

Which datasets are used for OD prediction in these papers?

BiST-IF uses HZMetro and NYC-TOD2018; DSTCN uses NYC-TOD2018, NYC-TOD2019 and HZMetro. Both report MAE and RMSE.

Which OD prediction baselines are compared?

Both compare with OD methods such as GEML and MPGCN and with spatio-temporal models such as STGCN, DGCN and STTN; BiST-IF also includes HIAM, and DSTCN includes HMOD, CMOD, STSGCN and TGCRN.

Fields, tasks, datasets and related search terms

Task

origin-destination (OD) demand prediction OD matrix forecasting travel demand forecasting taxi demand prediction metro passenger flow prediction spatio-temporal forecasting origin-destination (OD) flow prediction inter-station passenger flow forecasting taxi OD demand prediction incomplete OD data

Method

BiLSTM spatio-temporal graph learning Transformer spatial similarity GRU temporal convolution graph neural network for OD prediction OD delay correction arrival flow (Out-OD) bi-directional attention mutual information attention gated feature update

Datasets

NYC-TOD2018 NYC-TOD2019 HZMetro (Hangzhou metro) NYC yellow taxi NYC-TOD2018 (New York yellow taxi)

Compared with

GEML MPGCN HMOD CMOD STGCN STSGCN DGCN STTN TGCRN HIAM

中文

OD 需求预测 起讫点需求预测 OD 矩阵预测 出行需求预测 出租车需求预测 地铁客流预测 时空图学习 OD 流量预测 站间客流预测 出租车 OD 需求预测 OD 延迟校正

Fields (broad to narrow)

artificial intelligence machine learning data mining spatio-temporal data mining time series forecasting intelligent transportation systems smart cities urban computing traffic prediction travel demand forecasting origin-destination (OD) demand prediction dynamic spatio-temporal correlation modeling passenger flow prediction origin-destination (OD) prediction OD prediction with delayed data

Related areas

travel demand modeling public transit analytics metro systems ride-hailing demand taxi demand mobility on demand spatio-temporal graph neural networks Transformers for spatio-temporal data graph structure learning bidirectional LSTM temporal convolutional networks external factors rail transit smart card data (AFC) attention mechanisms recurrent neural networks missing data imputation data completion mutual information gating mechanisms

Applications

real-time traffic management dynamic vehicle dispatch fleet management resource allocation transport planning metro operations train scheduling crowd control in stations taxi dispatch

Related benchmarks (not used in this paper)

SHMetro NYC Citi Bike OD Chicago taxi OD Beijing subway

中文 (扩展)

出行需求预测 OD 需求 起讫点矩阵 客流预测 网约车需求预测 出租车调度 车辆调度 车队管理 时空图神经网络 图结构学习 双向LSTM 时序卷积网络 OD 预测 出行需求建模 公共交通分析 轨道交通 地铁 刷卡数据 网约车需求 数据补全 缺失数据填补 注意力机制 循环神经网络 互信息 门控机制 地铁运营 交通规划

Cite these papers

@article{gong2026dstcn,
  title        = {Exploiting dynamic spatio-temporal correlations for origin-destination demand prediction},
  author       = {Gong, Yongshun and Yu, Piao and Zhang, Xu and Zhang, Xinxin and Nie, Xiushan and Sun, Haoliang},
  journal      = {Expert Systems with Applications},
  year         = {2026},
  volume       = {299},
  pages        = {130095},
  doi          = {10.1016/j.eswa.2025.130095},
  issn         = {0957-4174},
  url          = {https://doi.org/10.1016/j.eswa.2025.130095}
}

@article{yu2025bistif,
  title        = {Enhancing origin–destination flow prediction via bi-directional spatio-temporal inference and interconnected feature evolution},
  author       = {Yu, Piao and Zhang, Xu and Gong, Yongshun and Zhang, Jian and Sun, Haoliang and Zhang, Junjie and Zhang, Xinxin and Yin, Yilong},
  journal      = {Expert Systems with Applications},
  year         = {2025},
  volume       = {264},
  pages        = {125679},
  doi          = {10.1016/j.eswa.2024.125679},
  issn         = {0957-4174},
  url          = {https://doi.org/10.1016/j.eswa.2024.125679}
}