==== RAD

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

Description

This repository contains code of Robust PCA and Robust Deep Autoencoder. Inspired by the Surus Project <https://github.com/Netflix/Surus>_ ( 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).

Functions

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.

Examples

• Robust Deep Autoencoder <http://nbviewer.jupyter.org/github/dlegor/RAD_Version_Python/blob/master/Notebook/Examples_and_Tests-Autoencoder.ipynb>_ (DEPRECATED)

• Robust PCA <http://nbviewer.jupyter.org/github/dlegor/rad/blob/master/notebooks/Examples_and_Tests-rPCA.ipynb>_

Installing

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:

.. code:: bash

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

Install the requirements

.. code:: bash

sudo pip install -r requirements.txt

or

.. code:: bash

conda install --file requirements.txt

Install the package

.. code:: bash

python setup.py install

References:

• Stable Principal Component Pursuit <https://arxiv.org/abs/1001.2363>_
• Robust Principal Component Analysis? <http://statweb.stanford.edu/~candes/papers/RobustPCA.pdf>_
• Anomaly Detection with Robust Deep Auto-encoders <https://www.kdd.org/kdd2017/papers/view/anomaly-detection-with-robust-deep-auto-encoders>_