Course Wiki

Logistic Regression

The classification model behind ML Zoomcamp's churn prediction project: a linear model squashed through the sigmoid function.

Module 3 predicts customer churn with logistic regression. The model computes a weighted sum of features like linear regression, then passes it through the sigmoid to produce a probability of the positive class. Training uses a likelihood-based loss rather than squared error.

The module’s practical work is feature engineering for categorical data (one-hot encoding), interpreting model output as churn risk, measuring feature importance from correlations and model weights, and using the trained model for targeted retention — scoring customers and acting on the top-risk segment.

Taught in

• Machine Learning Zoomcamp

  • Module 1: Introduction to Machine Learning
    • Model Selection Process
  • Module 3: Machine Learning for Classification
    • Logistic regression
    • Training logistic regression with Scikit-Learn
    • Model interpretation
    • Using the model
    • Summary ## Related concepts

• Classification Metrics

• Linear Regression

• Regularization

• Decision Trees

• Gradient Boosting