📊 Customer Churn Prediction - Multivariate Logistic Regression

End-to-end machine learning project with logistic regression to predict customer churn, using FastAPI for the backend and Streamlit for a modern, interactive frontend.

📋 Project Overview

This project predicts the likelihood of customer churn (service cancellation) for a telecommunications company. Using customer service, billing, and demographic data, it estimates whether a customer is at high risk of leaving.

• Cleaned dataset with over 7,000 customers
• More than 30 input features
• Integrated machine learning + API + frontend interface

🧪 Model Pipeline

• Feature selection and cleaning
• One-hot encoding for categorical variables
• Standardization of numerical features
• Logistic Regression training
• Model evaluation (Confusion Matrix, Accuracy, Precision, AUC)
• Model serialization with pickle

🚀 Backend - FastAPI

The model is served using a FastAPI backend for real-time predictions.

Running the API

1. Install dependencies:

   pip install fastapi uvicorn scikit-learn pandas numpy

2. Run the API:

   uvicorn fastAPI.main:app --reload

3. Access the Swagger docs at:

   • http://localhost:8000/docs

API Endpoint

• POST /predict
  • Input: JSON with customer attributes
  • Output: Churn probability and prediction (0 = no churn, 1 = churn)

Example Request:

{
  "Contract": 0,
  "Tech_Support": 1,
  "Tenure_Months": 10,
  "Online_Security": 2,
  "Internet_Service": 1,
  "Device_Protection": 0,
  "Payment_Method": 2,
  "Monthly_Charges": 85.90,
  "Online_Backup": 0,
  "Dependents": 0,
  "Streaming_TV": 1,
  "Streaming_Movies": 1
}

Example Response:

{
  "churn_probability": 0.742,
  "prediction": 1
}

🎨 Frontend - Streamlit

An interactive frontend built with Streamlit allows users to input customer data and view real-time predictions in an intuitive and styled interface.

Running the App

1. Install dependencies:

   pip install streamlit requests

2. Launch the app:

   streamlit run fastAPI\app.py

App Features

• User-friendly input forms with radio buttons, sliders, and dropdowns
• Clean two-column layout for better organization
• Real-time API communication
• Styled result output with probability, colors, and risk icons (🟢 / 🔴)
• Error handling and loading animations

Note

The code descriptions are in Portuguese 🇧🇷, although the variables and code are in English.

# Author Info

# LinkedIn: https://www.linkedin.com/in/profile-mariana-martins/
# GitHub: https://github.com/marianamartiyns
# Email: marianamartiyns@gmail.com