Dyn-GWN: Time-Series Forecasting using Time-varying Graphs with Applications to Finance and Traffic Prediction
Spatio-temporal modeling is an essential lens to understand many real-world phenomena from traffic to finance. There has been exciting work that explores spatio-temporal modeling with temporal graph convolutional networks. Often these methods assume that the spatial structure is static. We propose a...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Language: | English |
| Published: |
ACM|4th ACM International Conference on AI in Finance
2023
|
| Online Access: | https://hdl.handle.net/1721.1/153137 |
| _version_ | 1826199390218354688 |
|---|---|
| author | Ibrahim, Shibal Tell, Max Mazumder, Rahul |
| author2 | Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science |
| author_facet | Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science Ibrahim, Shibal Tell, Max Mazumder, Rahul |
| author_sort | Ibrahim, Shibal |
| collection | MIT |
| description | Spatio-temporal modeling is an essential lens to understand many real-world phenomena from traffic to finance. There has been exciting work that explores spatio-temporal modeling with temporal graph convolutional networks. Often these methods assume that the spatial structure is static. We propose a new model Dyn-GWN for spatio-temporal learning from time-varying graphs. Our model relies on a novel module called the Tensor Graph Convolutional Module (TGCM), which captures dynamic trends in graphs effectively in the time-varying graph representations. This module has two components: (i) it applies temporal dilated convolutions both on the time-varying graph adjacency space and the time-varying features. (ii) it aggregates the higher-level latent representations from both time-varying components through a proposed layer TGCL. Experiments demonstrate the efficacy of these model across time-series data from finance and traffic domains. Dyn-GWN can give up to better out-of-sample performance than prior methods that learn from time-varying graphs, e.g., EvolveGCN and TM-GCN. Interestingly, Dyn-GWN can be ∼ 300 × faster than EvolveGCN, which is the more competitive baseline from state-of-the-art models that cater to time-varying graphs. |
| first_indexed | 2024-09-23T11:19:29Z |
| format | Article |
| id | mit-1721.1/153137 |
| institution | Massachusetts Institute of Technology |
| language | English |
| last_indexed | 2024-09-23T11:19:29Z |
| publishDate | 2023 |
| publisher | ACM|4th ACM International Conference on AI in Finance |
| record_format | dspace |
| spelling | mit-1721.1/1531372024-01-11T19:37:45Z Dyn-GWN: Time-Series Forecasting using Time-varying Graphs with Applications to Finance and Traffic Prediction Ibrahim, Shibal Tell, Max Mazumder, Rahul Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science Center for Brains, Minds, and Machines Sloan School of Management Spatio-temporal modeling is an essential lens to understand many real-world phenomena from traffic to finance. There has been exciting work that explores spatio-temporal modeling with temporal graph convolutional networks. Often these methods assume that the spatial structure is static. We propose a new model Dyn-GWN for spatio-temporal learning from time-varying graphs. Our model relies on a novel module called the Tensor Graph Convolutional Module (TGCM), which captures dynamic trends in graphs effectively in the time-varying graph representations. This module has two components: (i) it applies temporal dilated convolutions both on the time-varying graph adjacency space and the time-varying features. (ii) it aggregates the higher-level latent representations from both time-varying components through a proposed layer TGCL. Experiments demonstrate the efficacy of these model across time-series data from finance and traffic domains. Dyn-GWN can give up to better out-of-sample performance than prior methods that learn from time-varying graphs, e.g., EvolveGCN and TM-GCN. Interestingly, Dyn-GWN can be ∼ 300 × faster than EvolveGCN, which is the more competitive baseline from state-of-the-art models that cater to time-varying graphs. 2023-12-12T13:56:10Z 2023-12-12T13:56:10Z 2023-11-27 2023-12-01T08:48:04Z Article http://purl.org/eprint/type/ConferencePaper 979-8-4007-0240-2 https://hdl.handle.net/1721.1/153137 Ibrahim, Shibal, Tell, Max and Mazumder, Rahul. 2023. "Dyn-GWN: Time-Series Forecasting using Time-varying Graphs with Applications to Finance and Traffic Prediction." PUBLISHER_CC en https://doi.org/10.1145/3604237.3626864 Creative Commons Attribution https://creativecommons.org/licenses/by/4.0/ The author(s) application/pdf ACM|4th ACM International Conference on AI in Finance Association for Computing Machinery |
| spellingShingle | Ibrahim, Shibal Tell, Max Mazumder, Rahul Dyn-GWN: Time-Series Forecasting using Time-varying Graphs with Applications to Finance and Traffic Prediction |
| title | Dyn-GWN: Time-Series Forecasting using Time-varying Graphs with Applications to Finance and Traffic Prediction |
| title_full | Dyn-GWN: Time-Series Forecasting using Time-varying Graphs with Applications to Finance and Traffic Prediction |
| title_fullStr | Dyn-GWN: Time-Series Forecasting using Time-varying Graphs with Applications to Finance and Traffic Prediction |
| title_full_unstemmed | Dyn-GWN: Time-Series Forecasting using Time-varying Graphs with Applications to Finance and Traffic Prediction |
| title_short | Dyn-GWN: Time-Series Forecasting using Time-varying Graphs with Applications to Finance and Traffic Prediction |
| title_sort | dyn gwn time series forecasting using time varying graphs with applications to finance and traffic prediction |
| url | https://hdl.handle.net/1721.1/153137 |
| work_keys_str_mv | AT ibrahimshibal dyngwntimeseriesforecastingusingtimevaryinggraphswithapplicationstofinanceandtrafficprediction AT tellmax dyngwntimeseriesforecastingusingtimevaryinggraphswithapplicationstofinanceandtrafficprediction AT mazumderrahul dyngwntimeseriesforecastingusingtimevaryinggraphswithapplicationstofinanceandtrafficprediction |