CSRSparseDataset

class lightautoml.dataset.np_pd_dataset.CSRSparseDataset(data, features=(), roles=None, task=None, **kwargs)[source]

Dataset that contains sparse features and np.ndarray targets.

Not implemented.

Convert to NumpyDataset.

property shape

Get size of 2d feature matrix.

__init__(data, features=(), roles=None, task=None, **kwargs)[source]

Create dataset from csr_matrix.

Parameters

data (Union[ndarray, csr_matrix, None]) – csr_matrix of features.
features (Union[Sequence[str], str, None]) – Features names.
roles (Union[Sequence[ColumnRole], ColumnRole, Dict[str, ColumnRole], None]) – Roles specifier.
task (Optional[Task]) – Task specifier.
**kwargs – Named attributes like target, group etc ..

Note

For different type of parameter feature there is different behavior:

list, should be same len as data.shape[1]

None - automatic set names like feat_0, feat_1 …

Prefix - automatic set names like Prefix_0, Prefix_1 …

For different type of parameter feature there is different behavior:

list, should be same len as data.shape[1].

None - automatic set NumericRole(np.float32).

ColumnRole - single role.

dict.

set_data(data, features=(), roles=None)[source]

Inplace set data, features, roles for empty dataset.

Parameters

data (Union[ndarray, csr_matrix]) – csr_matrix of features.
features (Union[Sequence[str], str, None]) – features names.
roles (Union[Sequence[ColumnRole], ColumnRole, Dict[str, ColumnRole], None]) – Roles specifier.

Note

For different type of parameter feature there is different behavior:

list, should be same len as data.shape[1]

None - automatic set names like feat_0, feat_1 …

Prefix - automatic set names like Prefix_0, Prefix_1 …

For different type of parameter feature there is different behavior:

list, should be same len as data.shape[1].

None - automatic set NumericRole(np.float32).

ColumnRole - single role.

dict.

static from_dataset(dataset)[source]

Convert dataset to sparse dataset.