Informer Forecasting with Directional Loss for High-Frequency Bitcoin Trading

Summary

The study tests the Informer forecasting architecture as a basis for automated strategies on high-frequency Bitcoin data. It trains variants with RMSE, Generalized Mean Absolute Directional Loss (GMADL), and quantile loss, then turns predicted future returns into trading strategies. Comparisons include Buy and Hold and two technical-indicator strategies. The evaluation covers 5-minute, 15-minute, and 30-minute data across six test periods.

Results depend on the loss function and sampling frequency. The quantile-loss model does not beat the benchmarks; RMSE performs worse at higher frequencies, while GMADL benefits from higher-frequency data. At 5-minute resolution, GMADL beats the other strategies in most test periods, according to the description. These findings suggest that a loss function emphasizing directional accuracy may matter for high-frequency trading forecasts. The evidence is limited to the reported Bitcoin periods and benchmarks, and the supplied summary gives no details on transaction costs, risk-adjusted returns, or performance outside the tested setup.

Key ideas

• The study converts Informer return forecasts into automated Bitcoin trading strategies.
• It compares RMSE, GMADL, and quantile loss against Buy and Hold and two indicator-based benchmarks.
• Quantile loss does not outperform the benchmarks in the reported evaluation.
• RMSE performance declines at higher data frequencies, while GMADL benefits from them.
• GMADL on 5-minute data beats the other tested strategies in most periods, within the study's evaluation.

Tags

# Informer in Algorithmic Investment Strategies on High Frequency Bitcoin Data


# Informer in Algorithmic Investment Strategies on High Frequency Bitcoin Data









The article investigates the usage of Informer architecture for building automated trading strategies for high frequency Bitcoin data. Three strategies using Informer model with different loss functions: Root Mean Squared Error (RMSE), Generalized Mean Absolute Directional Loss (GMADL) and Quantile loss, are proposed and evaluated against the Buy and Hold benchmark and two benchmark strategies based on technical indicators. The evaluation is conducted using data of various frequencies: 5 minute, 15 minute, and 30 minute intervals, over the 6 different periods. Although the Informer-based model with Quantile loss did not outperform the benchmark, two other models achieved better results. The performance of the model using RMSE loss worsens when used with higher frequency data while the model that uses novel GMADL loss function is benefiting from higher frequency data and when trained on 5 minute interval it beat all the other strategies on most of the testing periods. The primary contribution of this study is the application and assessment of the RMSE, GMADL, and Quantile loss functions with the Informer model to forecast future returns, subsequently using these forecasts to develop automated trading strategies. The research provides evidence that employing an Informer model trained with the GMADL loss function can result in superior trading outcomes compared to the buy-and-hold approach.

Shown in full with attribution under the source's licence. Licence: abstract CC0

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.