AnoViz: A Visual Inspection Tool of Anomalies in Multivariate Time Series

This is the implementation of a paper published in AAAI 2023 (Demonstration Track) [Paper] [Video] [Poster] [Live Demo]

Citation

@inproceedings{AnoVizAAAI,
  title={{AnoViz}: A Visual Inspection Tool of Anomalies in Multivariate Time Series},
  author={Trirat, Patara, and Nam, Youngeun, and Kim, Taeyoon and Lee, Jae-Gil},
  booktitle={Proceedings of the 37th AAAI Conference on Artificial Intelligence},
  pages={16489-16490},
  year={2023}
}

Basic Requirements

Web Client

• Nodejs v16+ with npm v8+ (See https://nodejs.org/en/)
• Firebase Web app (See https://firebase.google.com/docs/firestore/quickstart, you may use other real-time database services for the stream mode.)

API Server

• Python 3.9 with pip v21+ (See https://www.python.org/downloads/)
• Anoconda or Miniconda (See https://www.anaconda.com/products/distribution)

Installation & Running

Web Client

cd client
npm install
# the web application will be accessible at: http://localhost:3000
npm run start

For self-hosting:

npm run build

Then, upload everything in client/build/ to your web server.

API Server

cd server
conda create --name AnoViz python=3.9
conda activate AnoViz
pip install -r requirements.txt

For running the servers:

bash startup.sh

or

# to run "Anomaly Detector" server
python -u anomaly_detector.py
# to run "Computation API" server, the api server will be accessible at: http://localhost:5555
python -u api_server.py

For stream simulation:

# from scratch, i.e., first batch of data stream
python -u simulator.py --action run --collection data_streams --limit 360
# after the run command, use append instead
python -u simulator.py --action append --collection data_streams --limit 36

These commands are for Firebase-based services. You may modify the code depending on your use cases.

Acknowledgments

This work was supported by Mobile eXperience Business, Samsung Electronics Co., Ltd. (Real-time Service Incident Prediction Development).