gplearn

Why we included this project

Most ML libraries hand you a weight matrix and call it a day. gplearn takes a different route: it evolves actual mathematical expressions from your features using genetic programming, covering symbolic regression, classification, and feature transformation behind the familiar scikit-learn fit/predict interface. That means it drops into existing pipelines and grid searches without a rewrite. The payoff is a readable formula describing the relationship between inputs and outputs, which matters in engineering, finance, and scientific work where stakeholders want to see and audit the model. The SymbolicTransformer is also handy for automated feature engineering, since evolution can surface nonlinear combinations you would not have thought to build by hand. If you already know scikit-learn conventions, the learning curve is short, and the docs cover the evolutionary parameters well enough to tune them for your problem.