src.utils.preprocess

Module Contents

src.utils.preprocess.downcast(s, try_numeric=True, category=False)

Downcast a series to the lowest possible memory type.

Parameters

spd.Series

Series to downcast.

try_numeric: bool

If True it will try to read strings as numeric values.

category: bool

If True (string) objects will be cast as a category.

Returns

Downcasted series.

src.utils.preprocess.map_labels(series, kind='categorical', labels=None, backwards=False, **arg)

Map a Series values by the labels given.

Parameters

series: pd.Series

Series to map on.

kind: str

Indicator for kind of preprocess in series. With kind of {“categorical”, “ordinal”} the mapping is applied, otherwise not.

labels: dict

Defines with the mapping {key_0: value_0, etc.}.

arg:

Additional arguments.

Returns

pd.Series

Series with mapped values.

class src.utils.preprocess.MinMaxScaler(upper=1, lower=- 1)

MinMax Scaler like in sklearn, prevents total library import/dependency.

Initialize scaler with upper and lower boundary.

Parameters

upperfloat

upper boundary to scale to

lowerfloat

lower boundary to scale to

fit(self, X, y=None)

Get fit parameters.

Parameters

Xnp.array

preprocess to fit on

yNone

solely for consistency

Returns

self

instance with self.min, self.max defined.

transform(self, X)

Scales preprocess according to fitted parameters.

Parameters

Xnp.array

preprocess to scale

Returns

np.array

scaled preprocess

fit_transform(self, X, y=None)

Execute consecutively self.fit and self.transform.

Parameters

Xnp.array

preprocess to scale

yNone

solely for consistency

Returns

np.array

scaled preprocess

inverse_transform(self, X, y=None)

Scale back to original domain.

Parameters

Xnp.array

preprocess to scale

yNone

solely for consistency

Returns

np.array

scaled preprocess

src.utils.parser src.utils.snowflake