Implementation of Robust PCA and Robust Deep Autoencoder for Time Series to detect outliers.

This repository contains code of Robust PCA and Robust Deep Autoencoder. Inspired by the Surus Project ( from Netflix ), I made a version of Robust PCA for Time Series in order to compare the efficiency for the detection of outliers compared to Robust Deep Autoencoder (for Time Series).

The models are in two functions:

• AnomalyDetection_RPCA: implementaion of Robust PCA similar of the Netflix's propuest.
• (DEPRECATED) AnomalyDetection_AUTOENCODER: implementation and adaptation of paper "Anomaly Detection with Robust Deep Auto-encoders" Chong Zhou;Randy Paffenroth.

• Robust Deep Autoencoder (DEPRECATED)
• Robust PCA

RAD depends upon scikit-learn and numba .

Requirements:

• Python 3.6 or greater
• numpy
• scipy
• scikit-learn
• numba
• pandas

For a manual install get this package:

wget https://github.com/dlegor/rad/archive/master.zip
unzip master.zip
rm master.zip
cd rad

Install the requirements

sudo pip install -r requirements.txt

or

conda install --file requirements.txt

Install the package

python setup.py install

• Stable Principal Component Pursuit
• Robust Principal Component Analysis?
• Anomaly Detection with Robust Deep Auto-encoders