What it does This tool not only builds sklearn pipeline object, but also builds single main estimator or single preprocessing component. The output object type is based on the length of pipeline steps. When there is only one step (choose None for others), either a main estimator or preprocessor, the component is output directly instead of wrapping in a pipeline object.
A typical pipeline chains one or more preprocessing steps plus a final main estimator, for example, [VarianceThreshold, StandardScaler, SGDClassifier] which is composed of a feature selctor, a preprocessing scaler and a main estimator together. For more information, please refer to Scikit-learn pipeline Pipeline.
Pre-processing components come from sklearn.preprocessing, feature_selection, decomposition, kernel_approximation, cluster.FeatureAgglomeration, skrebate and more.
Final Estimator supports estimators from xgboost and many scikit-learn modules, including svm, linear_model, ensemble, naive_bayes, tree, neighbors and so on.
Custom estimators
Output