StreamAD

Anomaly detection for data streams/time series. Detectors process the univariate or multivariate data one by one to simulte a real-time scene.

Documentation

Installation

The stable version can be installed from PyPI:

pip install streamad

The development version can be installed from GitHub:

pip install git+https://github.com/Fengrui-Liu/StreamAD

Quick Start

Start once detection within 5 lines of code. You can find more example with visualization results here.

from streamad.util import StreamGenerator, UnivariateDS
from streamad.model import SpotDetector

ds = UnivariateDS()
stream = StreamGenerator(ds.data)
model = SpotDetector()

for x in stream.iter_item():
    score = model.fit_score(x)

Models

For univariate time series

If you want to detect multivarite time series with these models, you need to apply them on each feature separately.

Model Example	API Usage	Paper
KNNCAD	streamad.model.KNNDetector	Conformalized density- and distance-based anomaly detection in time-series data
SPOT	streamad.model.SpotDetector	Anomaly detection in streams with extreme value theory
Spectral Residual	streamad.model.SRDetector	Time-series anomaly detection service at microsoft
Z score	streamad.model.ZScoreDetector	Standard score
One-class SVM	streamad.model.OCSVMDetector	One-class SVM
MAD	streamad.model.MadDetector	Median absolute deviation
SARIMAX	streamad.model.SArimaDetector	Seasonal Arima Detector

For multivariate time series

These models are compatible with univariate time series.

Models Example	API Usage	Paper
xStream	streamad.model.xStramDetector	Xstream: outlier detection in feature-evolving data streams
RShash	streamad.model.RShashDetector	Subspace Outlier Detection in Linear Time with Randomized Hashing
HSTree	streamad.model.HSTreeDetector	Fast Anomaly Detection for Streaming Data
LODA	streamad.model.LodaDetector	Lightweight on-line detector of anomalies
RRCF	streamad.model.RrcfDetector	Robust random cut forest based anomaly detection on streams