🧠 Customer Churn Prediction using ANN

A complete end-to-end deep learning project built with an Artificial Neural Network (ANN) to predict customer churn — determining whether a customer is likely to leave a company or stay, based on their banking data.

---

🚀 Project Overview

This project aims to help businesses retain customers by predicting churn using historical data.

It uses the Churn_Modelling.csv dataset and includes all essential steps:

• Data preprocessing (encoding, scaling, splitting)
• Building and training an ANN model using Keras & TensorFlow
• Evaluating performance with metrics like accuracy and confusion matrix
• Saving trained models and preprocessing objects (.h5, .pkl)
• Simple prediction script (app.py) for real-time inference

This project can serve as a great portfolio example demonstrating your skills in Machine Learning, Deep Learning, and Model Deployment.

---

📁 Repository Structure

Customer-Churn-Prediction-ANN-model

--> Churn_Modelling.csv # Dataset

--> experiments.ipynb # Data preprocessing + model building notebook

--> prediction.ipynb # Example predictions using saved model

--> app.py # Script to run predictions

--> model.h5 # Trained ANN model

--> label_encoder_gender.pkl # Label encoder for Gender feature

--> onehot_encoder_geography.pkl # One-hot encoder for Geography

--> scaler.pkl # Feature scaler (StandardScaler or MinMaxScaler)

--> requirements.txt # List of dependencies

--> README.md # Project documentation

---

✅ Features

• End-to-end ANN model to predict customer churn

• Preprocessing using LabelEncoder, OneHotEncoder, and Scaler

• Model serialization (.h5 and .pkl files for reuse)

• Confusion matrix and accuracy evaluation

• Real-time prediction with sample customer input

• Easy retraining using Jupyter notebook

---

🔧 Setup & Installation

1. Clone the repository

   git clone https://github.com/Johan621/Customer-Churn-Prediction-ANN-model-.git
   cd Customer-Churn-Prediction-ANN-model-

2. Create and activate a virtual environment (optional but recommended)

python -m venv venv
venv\Scripts\activate   # For Windows

3. Install dependencies

pip install -r requirements.txt

4. Ensure the dataset Churn_Modelling.csv is present in the root directory.

5. To retrain the model, open experiments.ipynb and run all cells.

6. To make predictions, you can use:

   python app.py

   or open prediction.ipynb to test manually.

📊 Model Architecture

Input Layer: Number of neurons = number of features after encoding

Hidden Layers: Dense layers with ReLU activation

Output Layer: Single neuron with Sigmoid activation for binary classification

Optimizer: Adam

Loss Function: Binary Crossentropy

Insights:

Customers with low tenure, low credit score, or fewer products are more likely to churn.

Long-tenure, multi-product customers with higher balance tend to stay.

🎯 How to Use the Model

1. Load the saved model:

from tensorflow.keras.models import load_model
import joblib
model = load_model('model.h5')
scaler = joblib.load('scaler.pkl')

2. Prepare new customer data with the same features as training data.

3. Apply the same encoders and scaler.

4. Predict churn:

prediction = model.predict(new_data)
churn_status = 'Yes' if prediction > 0.5 else 'No'
print(churn_status)

🙋‍♂️ Author & Credits

Author: Johan621

Dataset: Banking Customer Churn Dataset (Churn_Modelling.csv) Frameworks: TensorFlow, Keras, Scikit-learn, Pandas, NumPy

📄 License

This project is licensed under the MIT License — see the LICENSE file for details.