eval_stats_clustering

Source code: sensai/evaluation/eval_stats/eval_stats_clustering.py

---

class ClusterLabelsEvalStats(labels: Sequence[int], noise_label: int, default_metrics: List[TMetric], additional_metrics: Optional[List[TMetric]] = None)

Bases: EvalStats[TMetric], ABC

NUM_CLUSTERS = 'numClusters'

AV_SIZE = 'averageClusterSize'

MEDIAN_SIZE = 'medianClusterSize'

STDDEV_SIZE = 'clusterSizeStd'

MIN_SIZE = 'minClusterSize'

MAX_SIZE = 'maxClusterSize'

NOISE_SIZE = 'noiseClusterSize'

get_distribution_summary() → Dict[str, float]

metrics_dict() → Dict[str, float]

Computes all metrics

Returns:

a dictionary mapping metric names to values

class ClusteringUnsupervisedMetric(name: Optional[str] = None, bounds: Optional[Tuple[float, float]] = None)

Bases: Metric[ClusteringUnsupervisedEvalStats], ABC

Parameters:

• name – the name of the metric; if None use the class’ name attribute

• bounds – the minimum and maximum values the metric can take on (or None if the bounds are not specified)

name: str

class RemovedNoiseUnsupervisedMetric(name: Optional[str] = None, bounds: Optional[Tuple[float, float]] = None)

Bases: ClusteringUnsupervisedMetric

Parameters:

• name – the name of the metric; if None use the class’ name attribute

• bounds – the minimum and maximum values the metric can take on (or None if the bounds are not specified)

worstValue = 0

compute_value_for_eval_stats(eval_stats: ClusteringUnsupervisedEvalStats) → float

abstract static compute_value(datapoints: ndarray, labels: Sequence[int])

name: str

class CalinskiHarabaszScore(name: Optional[str] = None, bounds: Optional[Tuple[float, float]] = None)

Bases: RemovedNoiseUnsupervisedMetric

Parameters:

• name – the name of the metric; if None use the class’ name attribute

• bounds – the minimum and maximum values the metric can take on (or None if the bounds are not specified)

name: str = 'CalinskiHarabaszScore'

static compute_value(datapoints: ndarray, labels: Sequence[int])

class DaviesBouldinScore(name: Optional[str] = None, bounds: Optional[Tuple[float, float]] = None)

Bases: RemovedNoiseUnsupervisedMetric

Parameters:

• name – the name of the metric; if None use the class’ name attribute

• bounds – the minimum and maximum values the metric can take on (or None if the bounds are not specified)

name: str = 'DaviesBouldinScore'

worstValue = 1

static compute_value(datapoints: ndarray, labels: Sequence[int])

class SilhouetteScore(name: Optional[str] = None, bounds: Optional[Tuple[float, float]] = None)

Bases: RemovedNoiseUnsupervisedMetric

Parameters:

• name – the name of the metric; if None use the class’ name attribute

• bounds – the minimum and maximum values the metric can take on (or None if the bounds are not specified)

name: str = 'SilhouetteScore'

worstValue = -1

static compute_value(datapoints: ndarray, labels: Sequence[int])

class ClusteringUnsupervisedEvalStats(datapoints: ndarray, labels: Sequence[int], noise_label=- 1, metrics: Optional[Sequence[ClusteringUnsupervisedMetric]] = None, additional_metrics: Optional[Sequence[ClusteringUnsupervisedMetric]] = None)

Bases: ClusterLabelsEvalStats[ClusteringUnsupervisedMetric]

Class containing methods to compute evaluation statistics of a clustering result

Parameters:

• datapoints – datapoints that were clustered

• labels – sequence of labels, usually the output of some clustering algorithm

• metrics – the metrics to compute. If None, will compute default metrics

• additional_metrics – the metrics to additionally compute

classmethod from_model(clustering_model: EuclideanClusterer)

clusterLabelsMask: ndarray

noiseLabelsMask: ndarray

class ClusteringSupervisedMetric(name: Optional[str] = None, bounds: Optional[Tuple[float, float]] = None)

Bases: Metric[ClusteringSupervisedEvalStats], ABC

Parameters:

• name – the name of the metric; if None use the class’ name attribute

• bounds – the minimum and maximum values the metric can take on (or None if the bounds are not specified)

name: str

class RemovedCommonNoiseSupervisedMetric(name: Optional[str] = None, bounds: Optional[Tuple[float, float]] = None)

Bases: ClusteringSupervisedMetric, ABC

Parameters:

• name – the name of the metric; if None use the class’ name attribute

• bounds – the minimum and maximum values the metric can take on (or None if the bounds are not specified)

worstValue = 0

compute_value_for_eval_stats(eval_stats: ClusteringSupervisedEvalStats) → float

abstract static compute_value(labels: Sequence[int], true_labels: Sequence[int])

name: str

class VMeasureScore(name: Optional[str] = None, bounds: Optional[Tuple[float, float]] = None)

Bases: RemovedCommonNoiseSupervisedMetric

Parameters:

• name – the name of the metric; if None use the class’ name attribute

• bounds – the minimum and maximum values the metric can take on (or None if the bounds are not specified)

name: str = 'VMeasureScore'

static compute_value(labels: Sequence[int], true_labels: Sequence[int])

class AdjustedRandScore(name: Optional[str] = None, bounds: Optional[Tuple[float, float]] = None)

Bases: RemovedCommonNoiseSupervisedMetric

Parameters:

• name – the name of the metric; if None use the class’ name attribute

• bounds – the minimum and maximum values the metric can take on (or None if the bounds are not specified)

name: str = 'AdjustedRandScore'

worstValue = -1

static compute_value(labels: Sequence[int], true_labels: Sequence[int])

class FowlkesMallowsScore(name: Optional[str] = None, bounds: Optional[Tuple[float, float]] = None)

Bases: RemovedCommonNoiseSupervisedMetric

Parameters:

• name – the name of the metric; if None use the class’ name attribute

• bounds – the minimum and maximum values the metric can take on (or None if the bounds are not specified)

name: str = 'FowlkesMallowsScore'

static compute_value(labels: Sequence[int], true_labels: Sequence[int])

class AdjustedMutualInfoScore(name: Optional[str] = None, bounds: Optional[Tuple[float, float]] = None)

Bases: RemovedCommonNoiseSupervisedMetric

Parameters:

• name – the name of the metric; if None use the class’ name attribute

• bounds – the minimum and maximum values the metric can take on (or None if the bounds are not specified)

name: str = 'AdjustedMutualInfoScore'

static compute_value(labels: Sequence[int], true_labels: Sequence[int])

class ClusteringSupervisedEvalStats(labels: Sequence[int], true_labels: Sequence[int], noise_label=- 1, metrics: Optional[Sequence[ClusteringSupervisedMetric]] = None, additional_metrics: Optional[Sequence[ClusteringSupervisedMetric]] = None)

Bases: ClusterLabelsEvalStats[ClusteringSupervisedMetric]

Class containing methods to compute evaluation statistics a clustering result based on ground truth clusters

Parameters:

• labels – sequence of labels, usually the output of some clustering algorithm

• true_labels – sequence of labels that represent the ground truth clusters

• metrics – the metrics to compute. If None, will compute default metrics

• additional_metrics – the metrics to additionally compute

classmethod from_model(clustering_model: EuclideanClusterer, true_labels: Sequence[int])

labels_with_removed_common_noise() → Tuple[ndarray, ndarray]

Returns:

tuple (labels, true_labels) where points classified as noise in true and predicted data were removed

clusterLabelsMask: ndarray

noiseLabelsMask: ndarray