CatBoost is an open-source gradient boosting library from Yandex that handles text-like category columns, such as city or colour, without you first converting them to numbers.
In plain words
Gradient boosting builds many decision trees in sequence, each fixing earlier mistakes. Many real tables have category columns, and most tools need you to turn those into numbers yourself. CatBoost does that conversion internally, using statistics about each category. It shuffles the rows first and uses an ordered scheme so the answer being predicted does not leak into the training features.
A quick example
Imagine predicting delivery times from a table with a column called "city" holding values like Paris or Lima. With many libraries you would first encode each city as numbers. With CatBoost you mark that column as categorical and it works out useful numbers for each city during training.
Where you'll meet it
- Tables of data with many category columns.
- Training on GPUs or on Apache Spark.
Related
Sources
- DocsCatBoost documentation, CatBoost · read 13 Oct 2026
- DocsTransforming categorical features to numerical features, CatBoost documentation · read 13 Oct 2026
- PaperCatBoost: unbiased boosting with categorical features, Prokhorenkova et al. (arXiv) · read 13 Oct 2026
- Docscatboost/catboost, CatBoost (GitHub) · read 13 Oct 2026
Checked by WhatIsAI against the sources above · 13 Oct 2026. Spotted an error? Report it.