feature: Add adaptive windowing for pattern change detection #10
Description
Implements three adaptive change point detection algorithms for real-time I/O frequency pattern analysis as part of my thesis work :
Algorithms Implemented:
- ADWIN (Adaptive Windowing): Statistical change detection using Hoeffding bounds with δ=0.05 (95% confidence)
- AV-CUSUM (Adaptive-Variance CUSUM): Bidirectional cumulative sum with 50-sample window, h_t=2σ threshold, k_t=0.5σ drift
- STPH (Self-Tuning Page-Hinkley): Adaptive running mean baseline with 10-sample window, h_t=2σ threshold, δ_t=0.5σ delta