A collection of anomaly detection methods (iid/point-based, graph and time series) including active learning for anomaly detection/discovery, bayesian rule-mining, description for diversity/explanation/interpretability. Analysis of incorporating label feedback with ensemble and tree-based detectors. Includes adversarial attacks with Graph Convolutional Network.

Algorithms for outlier and adversarial instance detection, concept drift and metrics.

AutoGBT is used for AutoML in a lifelong machine learning setting to classify large volume high cardinality data streams under concept-drift. AutoGBT was developed by a joint team ('autodidact.ai') from Flytxt, Indian Institute of Technology Delhi and CSIR-CEERI as a part of NIPS 2018 AutoML for Lifelong Machine Learning Challenge.

Algorithms for detecting changes from a data stream.

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

Repository for the AdaptiveRandomForest algorithm implemented in MOA 2016-04

unsupervised concept drift detection

My Java codes for the MOA framework. It includes implementations of FHDDM, FHDDMS, and MDDMs.

Concept Drift and Concept Shift Detection for Predictive Models

concept drift datasets edited to work with scikit-multiflow directly

a small example showing interactions between MLFlow and scikit-multiflow

Thanks to Latent Dirichlet Allocation and the ADWIN Algorithm, we realize topic modeling and concept drift detection among a corpus.

Concept Drift Detection Through Resampling - Algorithms Implementation

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

Queue-Based Resampling (QBR)

unsupervised concept drift detection with one-class classifiers

A classifier for heterogeneous concept drift inspired in the biologically memory model.

Code for my Master Thesis: How to detect and address changes in machine learning based data pipelines

Incremental Gaussian Mixture Network for Non-Stationary Environments

The implementation of the Diversity Pool algorithm, proposed in the paper "Diversity-Based Pool of Models for Dealing with Recurring Concepts" and presented at IJCNN '18

Machine Learning algorithms for MOA designed to cope with concept drift.

A Julia implementation of Stream Classification Algorithm Guided by Clustering – SCARGC

Landmark-based Feature Drift Detector

Efficient Multistream Classification using Direct DensIty Ratio Estimation

EDIST2: Error Distance Approach for Drift Detection and Monitoring

Repository for the StreamingRandomPatches algorithm implemented in MOA 2019.04

Adaptive REBAlancing (AREBA)

This project uses Time Series model (ARMA) to work with Concept Drift.