This repository contains the 4th position solution for the Pre-Bootcamp hackathon organised by Data Science Nigeria (DSN) on Zindi, from 8 August—22 August, 2020. (link to hackathon: https://zindi.africa/hackathons/dsn-pre-bootcamp-hackathon-expresso-churn-prediction-challenge).
To help Expresso to better serve their customers by understanding which customers are at risk of leaving.
To develop a predictive model that determines the likelihood for a customer to churn - to stop purchasing airtime and data from Expresso
Scikit learn Pandas Numpy Matplotlib Catboost Lightgbm Seaborn
Logloss
0.246701956963814
The solution was built on two models - Catboost and lightgbm, and a weighted average of them, with the averaged model achieving a good performance on the private leaderboard.