custom-anomaly-detectors

Overview

A tiny package containing some custom anomaly detectors. In the models.py source code there are 4 detectors:

1. Multivariate Gaussian based anomaly detector

   from pyad.models import MultivariateGaussian

2. Multivariate T Distribution based anomaly detector

   from pyad.models import MultivariateT

3. Simple independence anomaly detector

   from pyad.models import SimpleAnomalyDetector

4. Influenced Outlierness based anomaly detector

   from pyad.models import INFLO

Installation and Dependencies

This is 'almost' a Python package (I will continuously work on this). But to use the code right away, simply clone this Github repo:

git clone https://github.com/viethungha0610/custom-anomaly-detectors.git

Environment

No complicated dependencies here. The package should run in any basic Conda Python 3.3+ environment with numpy, scipy and scikit-learn. Although I would recommend running in a Python 3.6+ environment.

However, you can pip install these packages by using the command line in this directory, and then run:

conda create -n YOURENVNAME python=3.6

conda activate YOURENVNAME

pip install -r requirements.txt

Using the code

Open a Jupyter notebook, import the classes from the first section and get going right away.

Suggested test dataset

Link to Kaggle Credit Card Fraud Detection dataset: https://www.kaggle.com/mlg-ulb/creditcardfraud