Classification Metrics

Definitions

Single Label Classification measures the performance of a classification module.

Example Usage

Required data items: predicted_class, ground_truth_class

res = client.metrics.compute(
    Metric.SingleLabelClassification,
    args={
        "predicted_class": "quantitative_question",
        "ground_truth_class": "qualitative_question",
    },
)
print(res)

Example Output

{
  'classification_prediction': 'quantitative_question', 'classification_ground_truth': 'qualitative_question', 'classification_correct': False
}

and the aggregated values

{
  'accuracy': 0.8, 
  'balanced_accuracy': 0.83, 
  'precision': 0.83, 
  'recall': 0.83, 
  'f1': 0.8
}