Domain description (domain)

Description of a domain stores a list of features, class(es) and meta attribute descriptors. A domain descriptor is attached to all tables in Orange to assign names and types to the corresponding columns. Columns in the Orange.data.Table have the roles of attributes (features, independent variables), class(es) (targets, outcomes, dependent variables) and meta attributes; in parallel to that, the domain descriptor stores their corresponding descriptions in collections of variable descriptors of type Orange.data.Variable.

Domain descriptors are also stored in predictive models and other objects to facilitate automated conversions between domains, as described below.

Domains are most often constructed automatically when loading the data or wrapping the numpy arrays into Orange’s Table.

>>> from Orange.data import Table
>>> iris = Table("iris")
>>> iris.domain
[sepal length, sepal width, petal length, petal width | iris]
class Orange.data.Domain(attributes, class_vars=None, metas=None, source=None)[source]
attributes

A tuple of descriptors (instances of Orange.data.Variable) for attributes (features, independent variables).

>>> iris.domain.attributes
(ContinuousVariable('sepal length'), ContinuousVariable('sepal width'),
ContinuousVariable('petal length'), ContinuousVariable('petal width'))
class_var

Class variable if the domain has a single class; None otherwise.

>>> iris.domain.class_var
DiscreteVariable('iris')
class_vars

A tuple of descriptors for class attributes (outcomes, dependent variables).

>>> iris.domain.class_vars
(DiscreteVariable('iris'),)
variables

A list of attributes and class attributes (the concatenation of the above).

>>> iris.domain.variables
(ContinuousVariable('sepal length'), ContinuousVariable('sepal width'),
ContinuousVariable('petal length'), ContinuousVariable('petal width'),
DiscreteVariable('iris'))
metas

List of meta attributes.

anonymous

True if the domain was constructed when converting numpy array to Orange.data.Table. Such domains can be converted to and from other domains even if they consist of different variable descriptors for as long as their number and types match.

__init__(attributes, class_vars=None, metas=None, source=None)[source]

Initialize a new domain descriptor. Arguments give the features and the class attribute(s). They can be described by descriptors (instances of Variable), or by indices or names if the source domain is given.

Parameters:
  • attributes (list of Variable) – a list of attributes
  • class_vars (Variable or list of Variable) – target variable or a list of target variables
  • metas (list of Variable) – a list of meta attributes
  • source (Orange.data.Domain) – the source domain for attributes
Returns:

a new domain

Return type:

Domain

The following script constructs a domain with a discrete feature gender and continuous feature age, and a continuous target salary.

>>> from Orange.data import Domain, DiscreteVariable, ContinuousVariable
>>> domain = Domain([DiscreteVariable.make("gender"),
...                  ContinuousVariable.make("age")],
...                 ContinuousVariable.make("salary"))
>>> domain
[gender, age | salary]

This constructs a new domain with some features from the Iris data set and a new feature color.

>>> new_domain = Domain(["sepal length",
...                      "petal length",
...                      DiscreteVariable.make("color")],
...                     iris.domain.class_var,
...                     source=iris.domain)
>>> new_domain
[sepal length, petal length, color | iris]
classmethod from_numpy(X, Y=None, metas=None)[source]

Create a domain corresponding to the given numpy arrays. This method is usually invoked from Orange.data.Table.from_numpy().

All attributes are assumed to be continuous and are named “Feature <n>”. Target variables are discrete if the only two values are 0 and 1; otherwise they are continuous. Discrete targets are named “Class <n>” and continuous are named “Target <n>”. Domain is marked as anonymous, so data from any other domain of the same shape can be converted into this one and vice-versa.

Parameters:
  • X (numpy.ndarray) – 2-dimensional array with data
  • Y (numpy.ndarray or None) – 1- of 2- dimensional data for target
  • metas (numpy.ndarray or None) – meta attributes
Returns:

a new domain

Return type:

Domain

>>> import numpy as np
>>> from Orange.data import Domain
>>> X = np.arange(20, dtype=float).reshape(5, 4)
>>> Y = np.arange(5, dtype=int)
>>> domain = Domain.from_numpy(X, Y)
>>> domain
[Feature 1, Feature 2, Feature 3, Feature 4 | Class 1]
__getitem__(idx)[source]

Return a variable descriptor from the given argument, which can be a descriptor, index or name. If var is a descriptor, the function returns this same object.

Parameters:idx (int, str or Variable) – index, name or descriptor
Returns:an instance of Variable described by var
Return type:Variable
>>> iris.domain[1:3]
(ContinuousVariable('sepal width'), ContinuousVariable('petal length'))
__len__()[source]

The number of variables (features and class attributes).

__contains__(item)[source]

Return True if the item (str, int, Variable) is in the domain.

>>> "petal length" in iris.domain
True
>>> "age" in iris.domain
False
index(var)[source]

Return the index of the given variable or meta attribute, represented with an instance of Variable, int or str.

>>> iris.domain.index("petal length")
2
has_discrete_attributes(include_class=False, include_metas=False)[source]

Return True if domain has any discrete attributes. If include_class is set, the check includes the class attribute(s). If include_metas is set, the check includes the meta attributes.

>>> iris.domain.has_discrete_attributes()
False
>>> iris.domain.has_discrete_attributes(include_class=True)
True
has_continuous_attributes(include_class=False, include_metas=False)[source]

Return True if domain has any continuous attributes. If include_class is set, the check includes the class attribute(s). If include_metas is set, the check includes the meta attributes.

>>> iris.domain.has_continuous_attributes()
True

Domain conversion

Domain descriptors also convert data instances between different domains.

In a typical scenario, we may want to discretize some continuous data before inducing a model. Discretizers (Orange.preprocess) construct a new data table with attribute descriptors (Orange.data.variable), that include the corresponding functions for conversion from continuous to discrete values. The trained model stores this domain descriptor and uses it to convert instances from the original domain to the discretized one at prediction phase.

In general, instances are converted between domains as follows.

  • If the target attribute appears in the source domain, the value is copied; two attributes are considered the same if they have the same descriptor.
  • If the target attribute descriptor defines a function for value transformation, the value is transformed.
  • Otherwise, the value is marked as missing.

An exception to this rule are domains in which the anonymous flag is set. When the source or the target domain is anonymous, they match if they have the same number of variables and types. In this case, the data is copied without considering the attribute descriptors.