A Machine Learning Analysis of CFPB Complaints

Group Members: Nora Hajjar, Cecile Murray, Erika Tyagi

This repository contains our final project for CAPP 30255 - Advanced Machine Learning for Public Policy, taken in the Spring of 2020. We use a variety of machine learning techniques to analyze complaints from the Consumer Financial Protection Bureau's Consumer Complaint Database. The main techniques that we use include unsupervised topic modeling, shallow learning classification, deep learning classification, and natural language processing).

Repository Structure

Our repository is organized as follows:

• The explore directory contains exploratory data analysis
• The clustering directory contains the clustering and topic modeling
• The shallow directory contains the shallow learning classification pipeline
• The deep directory contains the deep learning classification pipeline

Specific instructions for running the analyses are provided within each directory.

All code contained in this repository was written by team members specifically for this project. PyTorch code to build the neural networks leveraged code written for assignments in this course.

Virtual Environment Setup

If you're using conda, you'll want to deactivate that first with conda deactivate.

Create the virtual environment. (Do this once)

python3 -mvenv venv

Next, activate it. Do this when you're working on the project. You'll want to deactivate when you're doing something else.

source venv/bin/activate

Install required packages (Do this the first time, and if you get a missing package error)

pip3 install -r requirements.txt

Keep requirements.txt up to date by updating the list of packages inside it:

pip3 freeze > requirements.txt