Kowope_Mart: This repository contains my solution for the 2020 Artificial Intelligence Qualification Bootcamp hackathon organized by Data Science Nigeria (DSN) on Zindi, from 9 September—3 October, 2020. (link to hackathon: https://zindi.africa/hackathons/dsn-ai-bootcamp-qualification-hackathon/leaderboard)..
Aim: To help Kowope Mart to identify customers who are worthy of credit card, lines and loans by identifying which customers are at risk of defaulting in payment of loan.
Objective: To develop a predictive model that determines the likelihood of a customer defaulting loan payment.
Packages: Scikit-learn, Numpy, Pandas, Matplotlib, Lightgbm, Xgboost, CatBoost, RandomizedSearchCV, StratifiedKFold.
Models: The solution was built on a stack of 4 models (LIghtGBM, CatBoost. Xgboost, Gradient Boosting Machines)..
Kowope Mart is a Nigerian-based retail company with a vision to provide quality goods, education and automobile services to its customers at affordable price and reduce if not eradicate charges on card payments and increase customer satisfaction with credit rewards that can be used within the Mall. To achieve this, the company has partnered with DSBank on co-branded credit card with additional functionality such that customers can request for loan, pay for goods even with zero-balance and then pay back within an agreed period of time. This innovative strategy has increased sales for the company. However, there has been recent cases of credit defaults and Kowope Mart will like to have a system that profiles customers who are worthy of the card with minimum if not zero risk of defaulting.
You have been employed as a Data Scientist to leverage Machine learning to predict customers who are likely to default or not.