skdatasets.utils.scores.scores_table

skdatasets.utils.scores.scores_table(scores: np.typing.ArrayLike, stds: np.typing.ArrayLike | None = None, *, datasets: Sequence[str], estimators: Sequence[str], nobs: int | None = None, greater_is_better: bool = True, method: Literal['average', 'min', 'max', 'dense', 'ordinal'] = 'min', significancy_level: float = 0, paired_test: bool = False, two_sided: bool = True, default_style: Literal['html', 'latex', None] = 'html', precision: int = 2, summary_rows: Sequence[Tuple[str, Callable[..., SummaryRow]]] = (('Average rank', <function average_rank>),)) → pd.io.formats.style.Styler[source]

Scores table.

Prints a table where each row represents a dataset and each column represents an estimator.

Parameters:

  • scores (array-like) – Matrix of scores where each column represents a model. Either the full matrix with all experiment results or the matrix with the mean scores can be passed.

  • stds (array-like, default=None) – Matrix of standard deviations where each column represents a model. If scores is the full matrix with all results this is automatically computed from it and should not be passed.

  • datasets (sequence of str) – List of dataset names.

  • estimators (sequence of str) – List of estimator names.

  • nobs (int) – Number of repetitions of the experiments. Used only for computing significances when scores is not the full matrix.

  • greater_is_better (boolean, default=True) – Whether a greater score is better (score) or worse (loss).

  • method ({'average', 'min', 'max', 'dense', 'ordinal'}, default='average') – Method used to solve ties.

  • significancy_level (float, default=0) – Significancy level for considerin a result significant. If nonzero, significancy is calculated using a t-test. In that case, if paired_test is True, scores should be the full matrix and a paired test is performed. Otherwise, the t-test assumes independence, and either scores should be the full matrix or nobs should be passed.

  • paired_test (bool, default=False) – Whether to perform a paired test or a test assuming independence. If True, scores should be the full matrix. Otherwise, either scores should be the full matrix or nobs should be passed.

  • two_sided (bool, default=True) – Whether to perform a two sided t-test or a one sided t-test.

  • default_style ({'html', 'latex', None}, default='html') – Default style for the table. Use None for no style. Note that the CSS classes and textual formatting are always set.

  • precision (int) – Number of decimals used for floating point numbers.

  • summary_rows (sequence) – List of (name, callable) tuples for additional summary rows. By default, the rank average is computed.

Returns:

table – Table of mean and standard deviation of each estimator-dataset pair. A ranking of estimators is also generated.

Return type:

array-like