This project was created to predict the dropout likelihood of bank customers for a kaggle challenge. In the Jupyter Notebook file data from the .csv files containing variables for bank customer behavior are converted into a Pandas DataFrame. I created features from those variables that improved prediction performance such as utilization rate and credit usage that improved model performance. Predictions were uploaded to kaggle matching their criteria for the prediction set.
The datasets were provided by Kaggle for training (train.csv) and developing model predictions (test.csv). Data contains various features on bank customers such as country of residence, credit score, number of accounts, and account balances.