A machine learning project that uses classification algorithms to predict customer attrition for a bank.
Banks must adjust their marketing efforts and provide products to clients whom are more likely to stay with a bank. This can save the bank a lot of time, money, and resources.
As a team, six members of the Baruch MLDS club along with myself did the following:
- Exploratory Data Analysis
- Data Encoding
- Training Multiple Classification Models
- Feature Engineering
- Model Evaluation
After saving the model, the team created a simple streamlit interface which takes in features like the client's age, months with a bank, total transaction amount (for the past 12 months), and the total revolving balance, to predict the likeliness of a customer's stay with the bank.
-
Clone the repository
-
Open the terminal and run
streamlit run app.py -
Have fun!
There are multiple improvements that can be done for this project:
- Find a solution to the class imbalance
- Custom Feature Engineering
- Train a LightGBM
- Collect more data
None of this would be possible without the awesome libraries Python comes with and most importantly, the team I guided and taught for the duration of the semester at Baruch College.