🤖 End-to-End Customer Churn Prediction Platform

Welcome to my Customer Churn Prediction Platform project! This project demonstrates a complete data science workflow — from raw data ingestion, transformation, and model training to deployment as an interactive web application.

➡️ Try the Live App Here!

🎯 Project Background & Objectives

Customer churn — when users stop subscribing or using a service — is a critical problem for subscription-based businesses. Losing customers not only reduces revenue but also increases the cost of acquiring replacements.

This project aims to build an automated system that can:

1. Identify high-risk customers before they churn.
2. Provide an interactive tool for business teams to perform "what-if" analysis on various customer profiles.

---

🏗️ System Architecture

This project is built using a modern data stack, separating each phase of the data lifecycle. The workflow is as follows:


1. Ingestion: Raw data is uploaded to Google Cloud Storage (GCS), acting as the data lake.
2. Warehousing: Data is loaded into Google BigQuery for structured storage and querying.
3. Transformation: dbt (data build tool) is used to clean, transform, and shape the raw data into analytics-ready models.
4. Modeling: A Jupyter notebook pulls the transformed data from BigQuery, trains a classification model using Scikit-learn, and saves it as a model artifact.
5. Deployment: An interactive web app is built using Streamlit, loading the trained model and serving predictions to end-users.

---

🛠️ Tech Stack

• Cloud Provider: Google Cloud Platform (GCP)
• Data Storage: Google Cloud Storage, Google BigQuery
• Data Transformation: dbt (data build tool)
• Machine Learning: Python, Pandas, Scikit-learn
• Web Application: Streamlit
• Orchestration & Automation (Local): Python Scripts, Jupyter Notebook

---

🚀 How to Run the Project Locally

Follow these steps to run the project on your local machine:

1. Clone the Repository

   git clone https://github.com/bestoism/ChurnPredictionPlatform.git
   cd ChurnPredictionPlatform

2. Install Dependencies
   Ensure Python 3.8+ and pip are installed.

   pip install -r requirements.txt

3. Set Up Google Cloud Authentication

   gcloud auth application-default login

4. Run dbt Transformations

   cd churn_analytics
   dbt run
   cd ..

5. Launch the Streamlit App

   streamlit run app.py

   The app will be available at http://localhost:8501.

---

📈 Results & Learnings

• The model achieved around 80% accuracy in predicting churn on the test dataset.
• Biggest Challenge: Handling data type inconsistencies between BigQuery, Pandas, and Scikit-learn — highlighting the importance of explicit and consistent data validation.
• This project reinforced the value of modular data architecture and using the right tool for the right task (e.g., dbt for transformations, Streamlit for interactivity).

---

Feel free to fork this project, open an issue, or contribute! 🚀

---

🚧 This project is under construction — stay tuned for more updates!