Client:
AvailableProduction:
Codelex LabDate:
August 27, 2023Category:
BrandingLocation:
Wehner Tunnel 6/8Project Details
Customer churn is a major challenge for telecom companies, as losing customers directly affects revenue and growth. I built the Customer Churn Prediction App to demonstrate how machine learning can predict whether a customer will stay or leave, helping businesses proactively retain clients. The project also serves as an educational tool to showcase the end-to-end ML workflow, from data preprocessing to deployment.
The app uses the Telco Customer Churn Dataset from IBM, containing both numerical and categorical features about customer demographics, account details, services, and charges. Preprocessing included handling missing values, encoding categorical variables, scaling numerical features, and addressing class imbalance using SMOTE. This ensured the dataset was clean, balanced, and ready for model training.
I trained and evaluated Random Forest and XGBoost models to predict churn. These models were chosen for their ability to handle complex relationships, mixed feature types, and high-dimensional data. Evaluation metrics included accuracy, precision, recall, F1-score, and AUC, ensuring the predictions were reliable and actionable. The models also provided insights into feature importance, highlighting key factors influencing customer churn.
The trained model was deployed using Streamlit, providing an intuitive interface for both single-customer prediction and batch predictions via CSV upload. Users receive clear predictions with probability-based visualizations, helping interpret results easily. The app demonstrates practical ML deployment, making it a strong portfolio project for interviews and showcasing the application of AI in real-world business problems.
The app predicts whether a telecom customer will churn (leave) or stay, based on their profile and service usage. Users can input data for a single customer or upload a CSV for batch predictions. It provides probability-based results with clear visualizations for easy interpretation.
This AI-powered web app uses a trained machine learning model to assess churn risk for telecom customers. It analyzes demographic, account, service, and billing features to generate predictions. The app helps businesses identify at-risk customers and take proactive retention measures.
Developed a strategy to predict telecom customer churn using machine learning, aiming to help companies retain clients and reduce revenue loss. Focused on creating an end-to-end ML solution with a user-friendly web interface.
Implemented the project using Python, Streamlit, and Scikit-learn/XGBoost. Preprocessed the Telco Customer Churn Dataset, handled class imbalance with SMOTE, trained models, and deployed them as a web app for single and batch predictions.
Conducted exploratory data analysis to understand customer behavior and feature importance. Evaluated model performance using accuracy, precision, recall, F1-score, and AUC to ensure reliable predictions for real-world application.
Delivered clear, probability-based predictions through Streamlit, with visualizations and insights for each customer. Provided actionable results highlighting which features contribute most to churn, making it easy for businesses to interpret and act on.
