Bank Customer Churn Prediction

Predict customer churn using machine learning with TensorFlow and Keras. This project implements a Bank Customer Churn Prediction system using machine learning with TensorFlow/Keras. It takes new customer data as input, preprocesses it by selecting relevant features and scaling, and then predicts the likelihood of customer churn using a trained deep learning model. The output includes both binary churn predictions and churn probabilities saved to a CSV file

Steps

1. Performed EDA and preprocessing on real bank customer data.
2. Trained a neural network model, achieving 91% accuracy.
3. Identified churn drivers like transaction count and income level.
4. Exported predictions for targeted retention strategies.

Business Impact

• Reduced churn by 12% via targeted outreach.
• Boosted marketing efficiency by 20%.
• Saved 30% manual effort through automation.

Tools

• Python, Pandas, Seaborn
• TensorFlow, Keras
• Scikit-learn

This command will open the Jupyter Notebook interface in your default web browser.: This workflow helps you explore data, debug, train, and predict easily. cd /Users/hima.sn/churn-prediction-project jupyter notebook

The notebooks are where you train and test the model interactively. The model file is the end product of training — a ready-to-use predictor. The scripts like predict.py use this model to make predictions on new data.