LabeledInstance#

class capymoa.instance.LabeledInstance[source]#

Bases: Instance

An Instance with a class label.

Most classification datastreams will automatically return instances for you with the class label and index. For example, the capymoa.datasets.ElectricityTiny dataset:

>>> from capymoa.datasets import ElectricityTiny
...
>>> from capymoa.instance import LabeledInstance
>>> stream = ElectricityTiny()
>>> instance: LabeledInstance = stream.next_instance()
>>> instance.y_label
'1'

The label and index are NOT the same. One is a human-readable string and the other is a integer representation of the class label. >>> instance.y_index 1 >>> instance.x array([0. , 0.056443, 0.439155, 0.003467, 0.422915, 0.414912])

__init__( schema: Schema, instance: InstanceExample | Tuple[ndarray[tuple[Any, ...], dtype[float64]], int], ) → None[source]#

Creates a new instance.

Its recommended that you prefer using from_array() or from_java_instance() to create instances, as they provide a more user-friendly interface.

Parameters:

schema – A schema that describes the datastream the instance belongs to.
instance – A vector of features (float values) or a Java instance.

Raises:

ValueError – If the given instance type is of an unsupported type.

classmethod from_array( schema: Schema, x: ndarray[tuple[Any, ...], dtype[float64]], y_index: int, ) → LabeledInstance[source]#

Creates a new labeled instance from a schema, feature vector, and class index.

This is useful in the rare cases you need to create custom labeled instances from scratch. In most cases, your datastream will automatically create instances for you.

>>> from capymoa.stream import Schema
...
>>> from capymoa.instance import LabeledInstance
>>> import numpy as np
>>> schema = Schema.from_custom(
...     ["f1", "f2", "class"],
...     target="class",
...     name="CustomDataset",
...     categories={"class": ["yes", "no"]}
... )
>>> x = np.array([0.1, 0.2])
>>> instance = LabeledInstance.from_array(schema, x, 0)
>>> instance
LabeledInstance(
    Schema(CustomDataset),
    x=[0.1 0.2],
    y_index=0,
    y_label='yes'
)
>>> instance.y_label
'yes'
>>> instance.java_instance.toString()
'0.1,0.2,yes,'

classmethod from_csv_row(schema: Schema, row: Sequence[str]) → Instance[source]#

Create an instance from a CSV row.

>>> from capymoa.stream import Schema
>>> from capymoa.instance import Instance
>>> schema = Schema.from_custom(
...     ["feature1", "feature2", "target"],
...     target="target",
...     categories={"feature2": ["A", "B"], "target": ["yes", "no"]},
...     name="classification-example"
... )
>>> row = ["1.0", "A", "yes"]
>>> instance = Instance.from_csv_row(schema, row)
>>> instance
Instance(
    Schema(classification-example),
    x=[1. 0.],
)
>>> instance.x
array([1., 0.])

Parameters:

schema – A schema providing the structure of each row. Like the header of the CSV.
row – A sequence of strings representing a CSV row.

Raises:

ValueError – If an attribute type is unsupported.

Returns:

A new Instance object.

classmethod from_java_instance( schema: Schema, java_instance: InstanceExample, ) → Instance[source]#

property java_instance: InstanceExample#: Returns a representation of the instance in Java for use in MOA. This method is for advanced users who want to directly interact with MOA’s Java API.

property schema: Schema#: Returns the schema of the instance and the stream it belongs to.

property x: ndarray[tuple[Any, ...], dtype[float64]]#

Returns a feature vector containing float values for the instance.

NaN values indicate missing features.

property y_index: int#

Returns the index of the class. It is useful for classification tasks as it provides a numeric representation of the class label, ranging from zero to the number of classes.

Values of -1 indicate a missing class label.

property y_label: str#: Returns the class label of the instance as a string.

LabeledInstance#

This Page