Benchmarking 16 deep learning time series forecasting architectures in financial scenarios released
Get up to speed
Adir Saly-Kaufmann, Kieran Wood, Jan-Peter Calliess, and Stefan Zohren have published a benchmark study covering 16 deep learning time series forecasting architectures in financial scenarios, which has been accepted by the Journal of Forecasting. The study tests architectures including Transformers, state space models (SSMs), xLSTM, and recurrent hybrid models on a 15-year daily futures dataset, using end-to-end volatility-targeted Sharpe ratio optimization as the benchmark task.
AI-generated from reports · 23 hr agoUpdated
Coverage timeline
Follow the coverage to understand different sides of the event.
- arXiv q-finHow Well Do Standard Time-Series Forecasting Benchmarks Transfer to Finance? A Large-Scale Study Covering Risk-Adjusted Performance
This paper tests 16 deep learning time-series forecasting architectures on a 15-year daily futures dataset, using end-to-end volatility-targeted Sharpe ratio optimization as the benchmark, covering Transformers, state space models (SSMs), xLSTM, and recurrent hybrid architectures.
Trend of this event
- Comparable range current
- 5
- Comparable range peak
- 926 Oct 15:00
- Change in last 24 hours
- –
Trend compares only entities observed continuously and completely, so the range may be smaller than the current heat count. Hover or tap the chart for hourly heat; use left/right arrow keys.