Bank Customer Churn Prediction using ANN

This project implements an Artificial Neural Network (ANN) to predict whether a bank customer is likely to leave (churn) based on various demographic and financial factors.

Project Overview

The model is built using Keras/TensorFlow and follows a standard deep learning workflow:

1. Data Loading: Importing the Churn Modeling dataset.
2. Preprocessing: Separating features and target, followed by Standard Scaling to normalize input data.
3. Architecture: A Sequential model with three hidden layers and a Sigmoid output for binary classification.
4. Evaluation: Monitoring accuracy on both training and test sets to check for overfitting.
5. Persistence: Saving the trained model and the scaler for future inference.

Technical Specifications

Model Architecture

The neural network is structured as follows:

• Input Layer: 8 input features.
• Hidden Layer 1: 6 neurons, ReLU activation.
• Hidden Layer 2: 4 neurons, ReLU activation.
• Hidden Layer 3: 2 neurons, ReLU activation.
• Output Layer: 1 neuron, Sigmoid activation (for probability output).

Hyperparameters

• Optimizer: Adam
• Loss Function: Binary Crossentropy
• Batch Size: 100
• Epochs: 50

Getting Started

Prerequisites

Ensure you have the following Python packages installed:

pip install pandas tensorflow scikit-learn