pd_provider

Pandas-backed unlabeled data provider.

class pysatl_cpd.data.providers.plain.pd_provider.PandasDataProvider(dataset, annotation)[source]

Bases: DataProvider[NumericArray, AnnotationT], Generic

Unlabeled pandas-backed provider for numeric feature arrays.

Parameters:

dataset (DataFrame) – Underlying pandas DataFrame.
annotation (TypeVar(AnnotationT, bound= UnlabeledTimeseriesAnnotation)) – Annotation for the dataset.

__init__(dataset, annotation)[source]

Parameters:

dataset (DataFrame)
annotation (AnnotationT)

Return type:

None

__iter__()[source]

Iterate over rows of the dataset.

Returns:: Iterator over scalar values for single-column data or row arrays for multivariate data.
Return type:: Iterator[ndarray[tuple[int, ...], dtype[double]]]

__len__()[source]

Return the number of rows in the dataset.

Returns:: Number of samples stored in the provider.
Return type:: int

property annotation: AnnotationT

Get annotation.

Returns:: The annotation object.
Return type:: annotation

property dataset: DataFrame

Get dataset copy.

Returns:: Copy of the underlying DataFrame.
Return type:: dataset

property columns: Sequence[str]

Get column names.

Returns:: List of column names.
Return type:: columns

select_columns(columns, *, rename_provider=False)[source]

Select subset of columns.

Parameters:

columns (Sequence[str]) – Column names to select.
rename_provider (bool) – If True, updates the provider name to reflect the selected columns.

Returns:

New provider with selected columns.

Return type:

PandasDataProvider[TypeVar(AnnotationT, bound= UnlabeledTimeseriesAnnotation)]

Raises:

ValueError – If no columns are selected or unknown columns are requested.

create_feature_column(*, name, mapping, rename_provider=False)[source]

Append a derived feature column computed row-wise.

Parameters:

name (str) – Name of the new feature column.
mapping (Callable[[Series], object]) – Callable applied to each row of the dataset.
rename_provider (bool) – If True, updates the provider name to reflect the new column.

Returns:

New provider with the appended feature column.

Return type:

PandasDataProvider[TypeVar(AnnotationT, bound= UnlabeledTimeseriesAnnotation)]

Raises:

ValueError – If the column name is empty or already exists.

cut(start, stop, *, annotation=None)[source]

Slice dataset by row indices.

Parameters:

start (int) – Start row index (inclusive).
stop (int) – Stop row index (inclusive).
annotation (Optional[TypeVar(AnnotationT, bound= UnlabeledTimeseriesAnnotation)]) – Optional annotation to use. If None, generates default.

Returns:

New provider with sliced data.

Return type:

PandasDataProvider[TypeVar(AnnotationT, bound= UnlabeledTimeseriesAnnotation)]

classmethod merge(providers, annotation_builder=None)[source]

Merge multiple providers.

Parameters:

providers (Sequence[Self]) – Sequence of providers to merge.
annotation_builder (Callable[[Sequence[TypeVar(AnnotationT, bound= UnlabeledTimeseriesAnnotation)]], TypeVar(AnnotationT, bound= UnlabeledTimeseriesAnnotation)] | None) – Optional function to build merged annotation.

Returns:

New merged provider.

Return type:

PandasDataProvider[TypeVar(AnnotationT, bound= UnlabeledTimeseriesAnnotation)]

Raises:

ValueError – If providers do not share the same columns and column order.