Method randomized_search

catboost/python-package/catboost/core.py:4522–4633

Randomized search on hyper parameters. After calling this method model is fitted and can be used, if not specified otherwise (refit=False). In contrast to grid_search, not all parameter values are tried out, but rather a fixed number of parameter settings is sampled

(self, param_distributions, X, y=None, cv=3, n_iter=10, partition_random_seed=0,
                          calc_cv_statistics=True, search_by_train_test_split=True, refit=True,
                          shuffle=True, stratified=None, train_size=0.8, verbose=True, plot=False, plot_file=None,
                          log_cout=None, log_cerr=None)

        )

    def randomized_search(self, param_distributions, X, y=None, cv=3, n_iter=10, partition_random_seed=0,
                          calc_cv_statistics=True, search_by_train_test_split=True, refit=True,
                          shuffle=True, stratified=None, train_size=0.8, verbose=True, plot=False, plot_file=None,
                          log_cout=None, log_cerr=None):
        """
        Randomized search on hyper parameters.
        After calling this method model is fitted and can be used, if not specified otherwise (refit=False).

        In contrast to grid_search, not all parameter values are tried out,
        but rather a fixed number of parameter settings is sampled from the specified distributions.
        The number of parameter settings that are tried is given by n_iter.

        Parameters
        ----------
        param_distributions: dict
            Dictionary with parameters names (string) as keys and distributions or lists of parameters to try.
            Distributions must provide a rvs method for sampling (such as those from scipy.stats.distributions).
            If a list is given, it is sampled uniformly.

        X: numpy.ndarray or pandas.DataFrame or polars.DataFrame or catboost.Pool
            Data to compute statistics on

        y: list or numpy.ndarray or pandas.DataFrame or pandas.Series or polars.DataFrame or polars.Series, optional (default=None)
            Labels of the training data.
            If not None, can be a single- or two- dimensional array with either:
              - numerical values - for regression (including multiregression), ranking and binary classification problems
              - class labels (boolean, integer or string) - for classification (including multiclassification) problems
            Use only if X is not catboost.Pool and does not point to a file.

        cv: int, cross-validation generator or an iterable, optional (default=None)
            Determines the cross-validation splitting strategy. Possible inputs for cv are:
            - None, to use the default 3-fold cross validation,
            - integer, to specify the number of folds in a (Stratified)KFold
            - one of the scikit-learn splitter classes
                (https://scikit-learn.org/stable/modules/classes.html#splitter-classes)
            - An iterable yielding (train, test) splits as arrays of indices.

        n_iter: int
            Number of parameter settings that are sampled.
            n_iter trades off runtime vs quality of the solution.

        partition_random_seed: int, optional (default=0)
            Use this as the seed value for random permutation of the data.
            Permutation is performed before splitting the data for cross validation.
            Each seed generates unique data splits.
            Used only when cv is None or int.

        search_by_train_test_split: bool, optional (default=True)
            If True, source dataset is splitted into train and test parts, models are trained
            on the train part and parameters are compared by loss function score on the test part.
            After that, if calc_cv_statistics=true, statistics on metrics are calculated
            using cross-validation using best parameters and the model is fitted with these parameters.

            If False, every iteration of grid search evaluates results on cross-validation.
            It is recommended to set parameter to True for large datasets, and to False for small datasets.

        calc_cv_statistics: bool, optional (default=True)
            The parameter determines whether quality should be estimated.