CatBoost

Summary

CatBoost (short for "Categorical Boosting") is a machine learning algorithm developed by Yandex, primarily designed for handling categorical data. It is an implementation of gradient boosting and is often used for classification and regression tasks.

Related Sections

Decision Trees
Classification Trees
Bias vs Variance
Encoding
Random Forest
AdaBoost
Gradient Boost

Key Principles

CatBoost uses a sophisticated version of target encoding to encode all discrete features within a dataset, then performs gradient boosting techniques to minimise the loss on sequentially generated weak learners to end up with an overall model.

Additional Resources

For a detailed explanation of CatBoost, I highly recommend watching Josh Starmer's video as he provides an excellent breakdown of the concept: Watch Video