Publicação
T2f
| dc.contributor.author | Sousa, João | |
| dc.contributor.author | Henriques, Roberto | |
| dc.contributor.institution | Information Management Research Center (MagIC) - NOVA Information Management School | |
| dc.contributor.institution | NOVA Information Management School (NOVA IMS) | |
| dc.contributor.pbl | Springer Science Business Media | |
| dc.date.accessioned | 2026-06-16T08:37:02Z | |
| dc.date.available | 2026-06-16T08:37:02Z | |
| dc.date.issued | 2026-06 | |
| dc.description | Sousa, J., & Henriques, R. (2026). T2f: Actor-critic reinforcement learning for time-series forecasting. Neural Computing and Applications, 38(12), Article 510. https://doi.org/10.1007/s00521-026-12209-6 | |
| dc.description.abstract | Time-series forecasting of multiple related sequences presents unique challenges due to the complex interplay between individual series characteristics and global patterns. We present T2f, a forecasting method combining ensemble learning with an actor-critic architecture based on the Twin Delayed Deep Deterministic algorithm (TD3). T2f balances local and global patterns through both its architecture and learning approaches, integrating transformer-based pattern recognition with reinforcement learning for dynamic model selection. Our method incorporates temporal attention mechanisms and context-aware error measurement, aligning forecasting objectives with practical decision-making priorities. Comprehensive ablation studies demonstrate that T2f’s components provide synergistic benefits: the TD3-based optimizer contributes 18.8% error reduction over static weighting, while temporal attention adds 8.0% improvement, with the integrated system outperforming simple ensemble baselines by over 20%. Experimental results across five diverse datasets indicate T2f reduced mean absolute error by over 30% compared to statistical models and achieved up to 40% better performance on context-weighted metrics than competing approaches. While specialized models occasionally outperformed T2f on highly regular patterns, it consistently showed superior adaptability to contextual weights with faster convergence, typically reaching near-optimal performance within 25 epochs compared to 40+ for alternative methods, particularly on datasets with complex temporal dynamics. | en |
| dc.description.version | publishersversion | |
| dc.description.version | published | |
| dc.format.extent | 35 | |
| dc.format.extent | 2762781 | |
| dc.identifier.doi | 10.1007/s00521-026-12209-6 | |
| dc.identifier.issn | 0941-0643 | |
| dc.identifier.other | PURE: 164809148 | |
| dc.identifier.other | PURE UUID: ed144499-9f15-4bfc-aabd-e56c163a97c0 | |
| dc.identifier.other | Scopus: 105041916833 | |
| dc.identifier.other | ORCID: /0000-0002-4862-8177/work/217889930 | |
| dc.identifier.uri | http://hdl.handle.net/10362/203794 | |
| dc.identifier.url | https://www.scopus.com/pages/publications/105041916833 | |
| dc.identifier.url | https://github.com/jfpsousa/t2f | |
| dc.identifier.url | https://doi.org/10.5281/zenodo.20718387 | |
| dc.language.iso | eng | |
| dc.peerreviewed | yes | |
| dc.relation | https://doi.org/10.54499/UID/04152/2025 | |
| dc.subject | Time-series forecasting | |
| dc.subject | Actor-critic methods | |
| dc.subject | Transformer architecture | |
| dc.subject | Reinforcement learning | |
| dc.subject | Ensemble learning | |
| dc.subject | Software | |
| dc.subject | Artificial Intelligence | |
| dc.title | T2f | en |
| dc.title.subtitle | Actor-critic reinforcement learning for time-series forecasting | en |
| dc.type | journal article | |
| degois.publication.issue | 12 | |
| degois.publication.title | Neural Computing and Applications | |
| degois.publication.volume | 38 | |
| dspace.entity.type | Publication | |
| rcaap.rights | openAccess |
Ficheiros
Principais
1 - 1 de 1