Supervised Autoencoders, Noise Augmentation, and Triple Barrier Labels

Summary

This study examines supervised autoencoders for financial forecasting and their effect on trading strategy performance. It combines learned representations with fractionally differentiated features and triple barrier labels, and tests whether adding noise during training changes risk-adjusted returns. The evaluation uses Bitcoin, Litecoin, and Ethereum data spanning January 2016 through April 2022, with Sharpe and Information Ratios as performance measures.

The reported findings suggest that a carefully balanced amount of noise augmentation and a suitable autoencoder bottleneck can improve strategy effectiveness. Too much noise or an excessively large bottleneck can reduce performance, so these settings require tuning. The excerpt does not provide specific ratio values, comparison baselines, or details of validation and trading costs. Its conclusions therefore describe results within the stated assets and sample period, rather than establishing that the approach will work across markets or live trading conditions.

Key ideas

• The study evaluates supervised autoencoders for financial forecasting and strategy performance.
• It combines fractionally differentiated features with triple barrier labels.
• Noise augmentation and bottleneck size affect measured risk-adjusted returns.
• The reported results favor balanced noise and bottleneck settings over excessive values.
• The excerpt does not give numerical outcomes or enough validation detail to assess broader applicability.

Tags

# 2411.12753


# Supervised Autoencoders with Fractionally Differentiated Features and Triple Barrier Labelling Enhance Predictions on Noisy Data









This paper investigates the enhancement of financial time series forecasting with the use of neural networks through supervised autoencoders (SAE), to improve investment strategy performance. Using the Sharpe and Information Ratios, it specifically examines the impact of noise augmentation and triple barrier labeling on risk-adjusted returns. The study focuses on Bitcoin, Litecoin, and Ethereum as the traded assets from January 1, 2016, to April 30, 2022. Findings indicate that supervised autoencoders, with balanced noise augmentation and bottleneck size, significantly boost strategy effectiveness. However, excessive noise and large bottleneck sizes can impair performance.

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.