• Qingyang Wu, 1069332
• Jinxuan Chen, 1075516
• Xiaotong Wang, 1132828
• Luyao Chen, 1266572
• Yuchen Luo, 1153247

Research Goal: Analysis to find the 100 best merchants for the Buy Now, Pay Later (BNPL) Firm to work with.

Timeline: The timeline for the research area is 02/2021 - 10/2022 Inclusive.

Instructions: To run the pipeline, please visit the scripts directory and run the files in order:

1. download.py: This downloads external postcodes and mean total income datasets into the data/external directory.
2. process_external_data.py: This script processes the "income" excel table and creates a postcode and SA2 lookup table. Processed data are stored in the data/curated directory.
3. join_table.py: This script conducts preliminary preprocessing and merges all datasets into a single data file data/curated/full_data/. Note that to run this script, please input the location where the data files are stored, e.g. python3 join_table.py "../data/tables/".
4. fraud_model.py: This script uses delta files to build a logistic regression model and predict whether other transactions are fraud. All fraud transactions are removed from full dataset and the final clean dataset is stored in data/curated/full_data_without_fraud/.
5. feature_engineering.py: This file summarises each merchant's data into several features and split the dataset into training and test sets.
6. model.py: This script preprocesses, trains and fits models on each of the 3 selected features that will be used in ranking merchants. The models utilise merchants' historical data (2021) to predict their "revenue", "number of consumers" and "number of transactions" in 2022.
7. ranking_system.py: This file calculates a ranking score for each merchant and finds top 100 merhants overall and top 10 merchants in each segment.

Notebook Info: Information regarding notebooks written for this project can be found in the README in the notebooks directory.