PyCaret builds on scikit-learn

PyCaret sits above scikit-learn as a higher-level experimentation layer. The project describes itself as a Python wrapper around existing machine learning libraries, and scikit-learn is the most central dependency in that stack for tabular classification and regression workflows.

How they fit together

PyCaret packages setup, preprocessing, model comparison, tuning, and evaluation into a small set of workflow functions and experiment classes. Underneath that interface, many of the estimators, transformers, and pipeline conventions come from scikit-learn. Teams that use PyCaret usually adopt it to reduce boilerplate around experiments rather than to replace the underlying estimator ecosystem.

What this means in practice

If a team is already comfortable with scikit-learn, PyCaret can accelerate prototype work and standardized benchmarking. If the team needs fine-grained control over pipeline internals, custom estimator composition, or low-level API access, scikit-learn remains the deeper foundation that PyCaret depends on.