Probability Density Functions (PDF) Likelihood Anomaly Detection - Interactive DEMOs (2D + 3D)

2D App

3D App

Setup:

1. Create or activate an existing Python >= 3.7.4, <= 3.9.2 environment.
2. Create a virtual environment using conda with sh setup_venv.sh in terminal (mac OS) or Command Line (Windows)

Note - built and tested using a Python >= 3.7.4, <= 3.9.2 base environment.

Running:

Repository

This repository contains two interactive DASH app dashboards. The dashboards demonstrate anomaly detection in 2D and 3D when considering joint probability distributions.

• src: Source folder containing the 2D and 3D dashboard code.

  • dashboard_2d:

    • utils:
      • kernel_density.py = Contains statsmodels.nonparametric.kernel_density.kdemultivariate class with one modification.
      • models.py = Contains both MLCV and BP11 KDE classes.
      • simulation.py = Contains functions to generate synthetic data.
    • main.py = Application is run from main.py. App configuration can be modified here.
    • figure_layout.py = static layouts for plotly graph objects.
    • inputs.py = Generates data and builds KDE estimates.
    • callbacks.py = Contains all callback functions that enable app user interaction.
    • layout.py = Contains all code that manages the app's HTML layout.

  • dashboard_3d: Folder structure is direct replica of dashboard_2d

• requirements.txt: .txt file that contains all of the required python libraries to run main.py.

• setup_venv.sh: shell script to create the virtual environment