import h2o
from h2o.model import MetricsBase
[docs]class H2OAnomalyDetectionModelMetrics(MetricsBase):
def _str_items_custom(self):
return [
"Anomaly Score: {}".format(self.mean_score()),
"Normalized Anomaly Score: {}".format(self.mean_normalized_score()),
]
[docs] def mean_score(self):
"""
Mean Anomaly Score. For Isolation Forest represents the average of all tree-path lengths.
:examples:
>>> from h2o.estimators.isolation_forest import H2OIsolationForestEstimator
>>> train = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/anomaly/ecg_discord_train.csv")
>>> test = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/anomaly/ecg_discord_test.csv")
>>> isofor_model = H2OIsolationForestEstimator(sample_size=5, ntrees=7)
>>> isofor_model.train(training_frame = train)
>>> perf = isofor_model.model_performance()
>>> perf.mean_score()
"""
if MetricsBase._has(self._metric_json, "mean_score"):
return self._metric_json["mean_score"]
return None
[docs] def mean_normalized_score(self):
"""
Mean Normalized Anomaly Score. For Isolation Forest - normalized average path length.
:examples:
>>> from h2o.estimators.isolation_forest import H2OIsolationForestEstimator
>>> train = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/anomaly/ecg_discord_train.csv")
>>> test = h2o.import_file("http://s3.amazonaws.com/h2o-public-test-data/smalldata/anomaly/ecg_discord_test.csv")
>>> isofor_model = H2OIsolationForestEstimator(sample_size=5, ntrees=7)
>>> isofor_model.train(training_frame = train)
>>> perf = isofor_model.model_performance()
>>> perf.mean_normalized_score()
"""
if MetricsBase._has(self._metric_json, "mean_normalized_score"):
return self._metric_json["mean_normalized_score"]
return None