Generic Buy Now, Pay Later Project

Introduction

---

A generic Buy Now, Pay Later (BNPL) firm has begun offering a new “Pay in 5 Installments” feature. Merchants are looking to boost their customer base by forming a partnership with this firm and in return, the BNPL firm gets a small percentage of revenue (take rate) to cover operating costs. The BNPL firm is looking for the top 100 merchants to onboard.

With this as the motivation, we have been able to formulate a Ranking Model, to score a Merchant based on their Financial capabilities, Customer Base, Fraud risk and Sustainability in the Market.

How to use

---

Pipeline

To ensure all required libraries are installed, run (preferably in Python 3.9) pip install -r requirements.txt

To run the pipeline python3 main.py --path "some/path/data" --output "some/output/dir" -option

• - -path: input folder path where all the tables (customer, merchant, transactions) are stored
• - -output: folder where the processed/cleaned tables are to be stored. Should have the same parent directory as '- -path', see example below
• -option (optional): -d: download files only -c: clean files only -p: process files only -m: model files only none: download, clean, process, and model

E.g., python3 main.py --path "/home/generic-buy-now-pay-later-project-group-3/data/tables" --output "/home/generic-buy-now-pay-later-project-group-3/data" -p

The following scripts get executed in order:

scripts
├── Main.py 
├── Download.py
├── Clean.py
├── Process.py
├── Pre_model.py
├── Model.py

To run the fraud detection and Ranking models, execute the following notebooks in order:

notebooks
├── Model_results.ipynb
├── Ranking_System_P1.ipynb
└── Ranking_System_P2_Functions.ipynb

Note:

• Notebooks with the prefix 'DO_NOT_RUN' are just helper notebooks used to understand the data and analyse trends, need not be run.
• In all the above notebooks, user needs to specify the path variable in the top cell. A note is included, just need to set the 'dir' variable to point to the data folder.
• The analysis summary notebooks are meant for demonstration purposes only and are not to be run.

Visualisations and graphs

The graphs for trends and model evaluations are included in the Model_results.ipynb and Ranking_System_P2_Functions.ipynb notebooks.

Contributors

• Prathyush P Rao (1102225)
• Harshita Soni (1138784)
• Chaitanya Raghuvanshi (117645)
• James Barro (1082092)
• Ruitong Wang (1118966)