CustomerChurnPrediction

You can access the live application here:

Customer Churn Prediction App

• Customer Churn Prediction: Predicts if a customer is likely to churn or not.
  • Web Interface: An interactive web application built with Streamlit for easy use.
  • Trained Model: Includes a pre-trained Keras model for immediate use.

Follow these instructions to get a copy of the project up and running on your local machine.

Prerequisites

You will need to have Python installed on your system. You can download it from the official Python website.

Installation

1. Clone the repository:

   git clone https://github.com/your-username/customer-churn-prediction.git
   cd customer-churn-prediction
   

2. Create and activate a virtual environment:

   python3 -m venv venv
   source venv/bin/activate
   

   Note: On Windows, the activation command is venv\Scripts\activate

3. Install the required libraries:

   pip install -r requirements.txt
   

To run the Streamlit application, execute the following command in your terminal:

streamlit run app.py

This will open a new tab in your web browser with the application running locally at http://localhost:8501.

• app.py: The main file for the Streamlit web application.
  • nn.ipynb: A Jupyter Notebook containing the code for training the neural network.
  • prediction.ipynb: A Jupyter Notebook demonstrating how to make predictions with the trained model.
  • Churn_Modelling.csv: The dataset used for training the model.
  • model.keras: The pre-trained Keras model.
  • requirements.txt: A list of all the Python libraries required for this project.
  • gender_le.pkl, geography_ohe.pkl, scaler.pkl: The saved encoders and scaler used for data preprocessing.

The model is an Artificial Neural Network (ANN) built with TensorFlow and Keras. The training process involves:

1. Data Preprocessing: Loading the data, handling categorical features, and scaling the data.
2. Model Building: Creating a sequential model with two hidden layers.
3. Training: Training the model with the preprocessed data and saving it as model.keras.

This project uses the "Churn Modelling" dataset, which is available on Kaggle.