Package TEES :: Package Evaluators :: Module AveragingMultiClassEvaluator :: Class AveragingMultiClassEvaluator

Class AveragingMultiClassEvaluator

Evaluator.Evaluator --+
                      |
                     AveragingMultiClassEvaluator

An evaluator for multiclass classification results, where an example can belong to one of several classes. For calculating averages over multiple classes, one of the classes, "neg"/1 is considered to be negative while the others are considered to be different types of positive instances.

Instance Methods

[hide private]

__init__(self, examples, predictions=None, classSet=None)

source code

compare(self, evaluation)
Compare overall performance between two sets of classified examples.

source code

getData(self)
Return the EvaluationData corresponding to the main evaluation criterion for this Evaluator.

source code

_queueUntypedUndirected(self, example, prediction)
All examples within the same majorId (same sentence) are put in queue.

source code

_processUntypedUndirectedQueue(self)
Determines the untyped undirected performance by merging example pairs.

source code

_calculate(self, examples, predictions)
The actual evaluation

source code

toStringConcise(self, indent='', title=None)
Evaluation results in a human readable string format source code

toDict(self)
Evaluation results in a computationally easy to process dictionary format

source code

Inherited from Evaluator.Evaluator: saveCSV

Class Methods

[hide private]

evaluate(cls, examples, predictions, classSet=None, outputFile=None, verbose=True)
Enables using this class without having to manually instantiate it

source code

threshold(cls, examples, predictions)

source code

Static Methods

[hide private]

Inherited from Evaluator.Evaluator: average, pool

Class Variables

[hide private]

type = 'multiclass'

Method Details

[hide private]

compare(self, evaluation)

source code

Compare overall performance between two sets of classified examples.

Overrides: Evaluator.Evaluator.compare: (inherited documentation)

getData(self)

source code

Return the EvaluationData corresponding to the main evaluation criterion for this Evaluator.

Overrides: Evaluator.Evaluator.getData: (inherited documentation)

_queueUntypedUndirected(self, example, prediction)

source code

All examples within the same majorId (same sentence) are put in queue. Once major id (sentence) changes, these examples are processed.

_processUntypedUndirectedQueue(self)

source code

Determines the untyped undirected performance by merging example pairs. This statistic is only meaningful for examples representing directed edges where two consecutive examples are the two directed edges between a pair of nodes.