Research map / Time series / Long-term forecasting
Diffusion Convolutional Recurrent Neural Network: Data-Driven Traffic Forecasting
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use directed bidirectional diffusion convolution rather than an undirected graph with ChebNet convolution (ablation) | DCRNN on METR-LA · METR-LA, 15 min; 30 min; 1 hour | MAE: 2.80 -> 2.77; 3.24 -> 3.15; 3.81 -> 3.60 | Helped |
| Remove diffusion convolution and use identity transition matrices (ablation) | DCRNN on METR-LA · METR-LA validation | validation error | Hurt |
| Use bidirectional rather than forward-only random walk convolution (ablation) | DCRNN on METR-LA · METR-LA validation | validation error | Helped |
| Replace directed graph modeling with an undirected graph and ChebNet convolution (ablation) | DCRNN on METR-LA · METR-LA | MAE: 2.77 -> 2.80; 3.15 -> 3.24; 3.60 -> 3.81 | Hurt |
| Replace the encoder-decoder sequence model with fixed-vector diffusion convolution layers (ablation) | DCRNN on METR-LA · METR-LA, multiple forecasting horizons | MAE | Hurt |
| Add an encoder-decoder sequence-to-sequence framework (ablation) | DCRNN on METR-LA · METR-LA, multiple forecasting horizons | MAE | Helped |
| Add scheduled sampling to sequence-to-sequence training (ablation) | DCRNN-SEQ on METR-LA · METR-LA, multiple forecasting horizons | MAE | Helped |
| Always feed model predictions as decoder inputs during multi-step training (ablation) | DCRNN on METR-LA · METR-LA, multiple forecasting horizons | MAE | Hurt |
| Vary diffusion filter width K (ablation) | DCRNN on METR-LA · METR-LA validation | validation error | Mixed |
| Vary the number of units per layer (ablation) | DCRNN on METR-LA · METR-LA validation | validation error | Mixed |
Closest papers by meaning
- Adaptive Graph Convolutional Recurrent Network for Traffic Forecasting (NeurIPS 2020 · similarity 0.91)
- Dynamic Spatial-Temporal Graph Convolutional Neural Networks for Traffic Forecasting (AAAI 2019 · similarity 0.91)
- Spatio-Temporal Graph Structure Learning for Traffic Forecasting (AAAI 2020 · similarity 0.91)
- TESTAM: A Time-Enhanced Spatio-Temporal Attention Model with Mixture of Experts (ICLR 2024 · similarity 0.87)
- Efficient Traffic Prediction Through Spatio-Temporal Distillation (AAAI 2025 · similarity 0.86)
- Spatio-Temporal Pivotal Graph Neural Networks for Traffic Flow Forecasting (AAAI 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.