API deployed on Render: https://churncustmodels-1.onrender.com
Web app deployed on Streamlit: https://churncustomermodels.streamlit.app/
Please be aware that Render free instance will spin down with inactivity, which can delay requests by 50 seconds or more. As a result, the web application may experience delays in retrieving results from the API.
This project focuses on churn customer detection using machine learning techniques and web application development. Below are the main stages of the project:
🧹 Data Cleaning and Preprocessing: Cleaned and engineered features to ensure high-quality input data for model training.
🤖 Model Training: Tuned multiple models, including XGBoost, Random Forest, SVM, Decision Tree, as well as ensemble methods like Voting Classifier and Stacking Classifier, focusing on optimizing recall.
🌐 API Deployment: Hosted the models on Render, making them accessible via an API for easy integration.
🖥️ Web App Development: Built a dashboard with Streamlit to visualize customer data, model predictions, and automated AI-generated explanations, along with personalized retention emails.
🔗 Llama 3.1 Integration: Integrated Llama 3.1 using Groq to enhance the generation of predictive explanations and email content.