Unsupervised learning for illicit activity
Docker container on VM
Credit to @syoh for setting up the initial Docker image and for help setting up the computational environment. See the GitHub repo https://github.com/dddlab/reproducibility-demo/tree/prep-for-binder for a demo on spinning up Docker containers that work with Binder.
- Start a Jupyter Notebook environment with
- The Docker image will be created from
Dockerfileand includes the necessary packages to run Jupyter Book
setup.shwill download a utility to create a password and encryption keys for your Jupyter notebook
Goals of the project
This project works with the latest trade mis-invoicing estimates of the United Nations Economic Commission for Africa: Lépissier, Alice, Davis, William, & Ibrahim, Gamal. (2019). Trade Mis-Invoicing Dataset (Version 1).
While generating estimates of the dollar value of illicit trade has been helpful to shed light on the severity of the problem, the next step in the analysis is to further understand the nature of the illicit activity in terms of its origins, destinations, and sectors.
Therefore, the goal of this project is to extract meaningful insights on illicit trade using unsupervised machine learning techniques. By doing so, I can identify analytically relevant categories and dimensions of variation, in order to generate hypotheses and guide further work.
This project will apply the following techniques to the data:
- Dimension reduction using Principal Components Analysis (PCA)
- Graph analysis
A Jupyter Book write-up of this project is available at https://alicelepissier.com/jupyter-book-IFF/.