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

Paper: arXiv 2411.12753

Abstract

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.

Complexity vs Empirical Score

• Math Complexity: 7.5/10
• Empirical Rigor: 6.0/10 (capped from 8.0: transaction costs only partially modelled; see penalty flags)
• Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced mathematical concepts such as fractional differentiation, binomial series expansions, and statistical stationarity tests, while also featuring a rigorous empirical design with detailed backtesting methodology, risk-adjusted metrics, and specific algorithmic implementations.

Research Flowchart

  flowchart TD
  A["Research Goal: Enhance financial time series forecasting"] --> B["Data Collection<br>Bitcoin, Litecoin, Ethereum<br>Jan 2016 - Apr 2022"]
  B --> C["Preprocessing<br>Fractional Differentiation & Noise Augmentation"]
  C --> D["Labeling<br>Triple Barrier Method"]
  D --> E["Modeling<br>Supervised Autoencoders<br>Bottleneck Size Optimization"]
  E --> F["Performance Evaluation<br>Sharpe & Information Ratios"]
  F --> G["Key Findings<br>Balanced noise & bottleneck size improve returns<br>Excessive noise impairs performance"]

Cited References (1667)

1. HYPERTEXT ‘87: Proceedings of the ACM Conference on Hypertext (1987) — link
2. HYPERTEXT ‘89: Proceedings of the Second Annual ACM Conference on Hypertext (1989) — link
3. ECHT ‘90: Proceedings of the European Conference on Hypertext : Hypertext: Concepts, Systems and Applications (1990)
4. HYPERTEXT ‘91: Proceedings of the Third Annual ACM Conference on Hypertext (1991) — link
5. ECHT ‘92: Proceedings of the ACM Conference on Hypertext (1992) — link
6. HYPERTEXT ‘93: Proceedings of the Fifth ACM Conference on Hypertext (1993) — link
7. ECHT94: Proceedings of the 1994 ACM European Conference on Hypermedia Technology (1994) — link
8. HYPERTEXT ‘96: Proceedings of the Seventh ACM Conference on Hypertext : Docuverse Takes Form (1996) — link

0 of 1667 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.