Sparsity-enhanced wavelet transform with dynamic thresholding for financial time series denoising: Peter Tettey Yamak et al.
Researchers present SWiFTS-D, a new method for cleaning noisy financial data using machine learning and wavelet transforms. SWiFTS-D combines Temporal Convolutional Networks (TCN) for dynamic thresholding and Elastic Net regularization for feature selection. It adapts to market volatility and improves signal-to-noise ratios by 14.5 to 54.7 percent. The approach reveals clearer patterns in assets like Bitcoin, Microsoft stock, and crude oil. However, details on datasets and real-world scalability are not addressed.
What it examines
This paper introduces SWiFTS-D, a new method for cleaning noisy financial time series data. It combines wavelet transforms, Temporal Convolutional Networks for dynamic thresholding, and Elastic Net regularization to adaptively remove noise and keep important patterns, improving analysis and forecasting in volatile markets.
What it concludes
SWiFTS-D outperforms traditional denoising methods, boosting signal quality and accuracy across various financial datasets. Its applications include algorithmic trading, risk management, and economic trend analysis. Future work may explore broader datasets and further optimization, making it a valuable tool for financial technology and analytics.
Evidence objects
Researchers unveil SWiFTS-D, a novel method blending machine learning and wavelet transforms, which dynamically adapts to market volatility and outperforms traditional denoising for assets like Bitcoin, Ethereum, EUR/USD, Microsoft, and crude oil.
key_findings bullet 1 · key_findings · validation V0
SWiFTS-D leverages a Temporal Convolutional Network for real-time thresholding and Elastic Net regularization for feature selection, boosting signal-to-noise ratios by 14.5--54.7% and peak signal-to-noise ratios by 13.3--55.5%.
key_findings bullet 2 · key_findings · validation V0
Despite its promise in extracting clearer financial patterns and aiding trading decisions, the study lacks details on datasets, computational costs, and real-world scalability, calling for further validation in live trading environments.
key_findings bullet 3 · key_findings · validation V0
SWiFTS-D uniquely integrates Temporal Convolutional Network-driven dynamic thresholding and Elastic Net regularization within a wavelet framework for financial time series denoising. Its adaptive thresholding and sparsity-enhanced feature selection are novel, yielding significant signal-to-noise improvements. This compelling method advances AI-driven quantitative finance, especially algorithmic trading and risk management.
key_findings bullet 4 · key_findings · validation V0
Raw abstract and provenance
- … applications in algorithmic trading, risk management, and economic trend analysis. … We plan to address it in future work by integrating explainable AI techniques. …
Source row: 1802 · abstract type: snippet