The Tornado 🌪️ framework, designed and implemented for adaptive online learning and data stream mining in Python.

Algorithms for detecting changes from a data stream.

An approach to apply concept drifts and ADWIN on the streams event time to reason about the progress of watermarks

Code for testing Concept drift techniques on a real word dataset on a hexapod robot

Салон штор

This project implements an AutoML-driven adaptive drift detection framework for COVID-19 forecasting, leveraging ADWIN & DDM to dynamically adjust models in response to concept drift. It enhances predictive accuracy by intelligently retraining models, ensuring robust time-series forecasting under evolving data conditions.

Streaming anomaly detection in Rust - detectors, calibration, SOC triage. Powers eBPFsentinel

15. Višestruko / panel sečenje sliding window stability, expanding window, block bootstrap, stratified by era, cohort split, concept-drift detectors (ADWIN, DDM, Page-Hinkley)

Memory-bounded streaming ensemble (Online Bagging + ADWIN + Leveraged Bagging) with a Hoeffding-Tree champion, including a documented post-mortem on a MOA Java-heap memory-measurement bug and its fix.

ML model monitoring is not just drift detection. This repo benchmarks PSI, KS, MMD, and ADWIN across real failure scenarios like data drift, label shift, and model decay in production.

Streaming, drift-aware log anomaly detection with conformal calibration; reproducible, CI-tested Python and strict provenance.

Online peak-hour boundary detection and concept-drift analysis on Salvador bus-boarding streams (CapyMOA ADWIN/Page-Hinkley + GTFS schedule alignment). 6-person group project.