Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Informer in Algorithmic Investment Strategies on High Frequency Bitcoin Data

Filip Stefaniuk, Robert Ślepaczuk

TL;DR

This study formulates automated BTC trading strategies by applying the Informer Transformer to forecast returns under three loss functions: RMSE, GMADL, and Quantile. It comprehensively benchmarks Informer-based strategies against Buy-and-Hold and MACD/RSI baselines across 5-, 15-, and 30-minute data within rolling 2-year in-sample and 6-month out-of-sample windows, using a suite of performance metrics and sensitivity analyses. The results show that the GMADL-trained Informer strategy, especially at 5-minute frequency, consistently outperforms the Buy-and-Hold baseline and many technical-indicator strategies, while the Quantile variant underperforms. The key contribution is demonstrating that training Informer with GMADL can yield superior trading outcomes, highlighting the loss-function choice as a critical design factor for ML-driven trading systems, and providing an open-source framework for reproducible strategy comparison.

Abstract

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.

Informer in Algorithmic Investment Strategies on High Frequency Bitcoin Data

TL;DR

This study formulates automated BTC trading strategies by applying the Informer Transformer to forecast returns under three loss functions: RMSE, GMADL, and Quantile. It comprehensively benchmarks Informer-based strategies against Buy-and-Hold and MACD/RSI baselines across 5-, 15-, and 30-minute data within rolling 2-year in-sample and 6-month out-of-sample windows, using a suite of performance metrics and sensitivity analyses. The results show that the GMADL-trained Informer strategy, especially at 5-minute frequency, consistently outperforms the Buy-and-Hold baseline and many technical-indicator strategies, while the Quantile variant underperforms. The key contribution is demonstrating that training Informer with GMADL can yield superior trading outcomes, highlighting the loss-function choice as a critical design factor for ML-driven trading systems, and providing an open-source framework for reproducible strategy comparison.

Abstract

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.

Paper Structure

This paper contains 55 sections, 39 equations, 17 figures, 19 tables.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Efficient Market Hypothesis
  4. Early Statistical Approach
  5. Early Machine Learning Systems
  6. Modern Machine Learning Systems
  7. Data
  8. Additional Data
  9. The Cboe Volatility Index (VIX Index)
  10. The Federal Funds effective rates
  11. Crypto Fear/Greed index
  12. Data Windows
  13. Methodology
  14. Evaluation Framework
  15. Annualized Return Compounded (ARC)
  16. ...and 40 more sections

Figures (17)

  • Figure 1: Price of the BTC/USDT
  • Figure 2: Cboe Volatility Index (VIX)
  • Figure 3: Note: The Federal Funds effective rates
  • Figure 4: The Crypto Fear/Greed Index
  • Figure 5: Rolling data windows
  • ...and 12 more figures