FHR-SWD Reproducibility Package: Functional Hurst-Regularised Stationary Wavelet Denoising for Fractal-Structured Big Data Streams

This dataset accompanies the manuscript "Functional Hurst-Regularised Stationary Wavelet Denoising via Total Variation-Penalised B-Spline Estimation for Fractal-Structured Big Data Streams" submitted to Big Data Mining and Analytics (BDMA). The package contains all synthetic experimental results, statistical test outputs, runtime measurements, and the complete Python pipeline used to generate them. Data are synthetic fractional Gaussian noise (fGn) signals generated using the Davies-Harte method across three Hurst regime configurations (A: single transition H=0.8→0.2; B: multiple transitions H=0.3→0.7→0.3; C: stationary H=0.5) at four input SNR levels (0, 5, 10, 20 dB), with N_MC=100 Monte Carlo replications at T=16,384 samples. Six denoising methods are compared: FHR-SWD (proposed), Standard-SWT, SureShrink, EMD-IT, Raw-DFA-SWT, and Ridge-B-spline. The package includes the Google Colab pipeline notebook, a standalone figure reproduction script, and the complete LaTeX manuscript.