BAD - Benchmarking Anomaly Detection

The BAD framework is a distributed framework for benchmarking unsupervised anomaly detection algorithms.

For details, please refer to the official documentation.

Installation

BAD can be easily installed via pip with the command:

pip install bad-framework

this installs the bad command-line interface.

Example usage

Before running experiments, you need to start the BAD server processes:

bad server-start

Then, you can run a simple experiment with:

bad run -c lof -d shuttle

this executes the famous Local Outlier Factor (LOF) algorithm on the shuttle dataset.

By default, results are stored in the file ./bad_out.csv.

The output file contains execution times, hyperparameter settings and evaluation metrics for all executed experiments.

The output file can be easily plotted with any graphing library.

Please refer to the official documentation for a complete command line reference.

Copyright © 2020 Sivam Pasupathipillai - sivam.pasupathipillai@gmail.com.

All rights reserved.