cluster_completeness#

er_evaluation.metrics.cluster_completeness(prediction, reference)[source]#

Cluster completeness score (based on conditional entropy)

This wraps scikit-learn’s completeness score function.

Parameters:

  • prediction (Series) – membership vector for predicted clusters, i.e. a pandas Series indexed by mention ids and with values representing predicted cluster assignment.

  • reference (Series) – membership vector for reference (true) clusters, i.e. a pandas Series indexed by mention ids and with values representing reference cluster assignment.

Returns:

completeness score

Return type:

float

Notes

  • The prediction and reference membership vectors are inner joined before this metric is computed.

  • NA values are dropped from membership vectors prior to computing the metric.