Wavelet Transforms in Financial Time Series Analysis: A Review on Stock Price Prediction

Abstract

Introduction: Stock price prediction is notoriously tricky due to data complexities such as non-stationarity and noise.

Methods: This literature survey was conducted in order to systematically evaluate how wavelet transforms enhance forecasting by improving data preprocessing for machine learning models.

Results: The review of hybrid wavelet-ML architectures (e.g., LSTMs, Transformers) finds that they consistently outperform standalone models by leveraging wavelets' powerful denoising and multi-scale feature-extraction capabilities. Despite their effectiveness, persistent challenges include parameter optimization, model interpretability, and overfitting.

Conclusion: Future research priorities should include adaptive wavelet techniques and the development of more robust, transparent hybrid systems for practical financial applications.