def get(self):
n_classes = len([i for i in self.cm.sum_row.values() if i != 0])
n_clusters = len([i for i in self.cm.sum_col.values() if i != 0])
if (n_classes == n_clusters == 1) or (n_classes == n_clusters == 0):
return 1.0
mutual_info_score = metrics.MutualInfo(self.cm).get()
expected_mutual_info_score = metrics.expected_mutual_info(self.cm)
entropy_true = _entropy(cm=self.cm, y_true=True)
entropy_pred = _entropy(cm=self.cm, y_true=False)
normalizer = self._generalized_average(entropy_true, entropy_pred)
denominator = normalizer - expected_mutual_info_score
if denominator < 0:
denominator = min(denominator, -np.finfo("float64").eps)
else:
denominator = max(denominator, np.finfo("float64").eps)
adjusted_mutual_info_score = (
mutual_info_score - expected_mutual_info_score
) / denominator
return adjusted_mutual_info_score
def get(self):
n_classes = len([i for i in self.cm.sum_row.values() if i != 0])
n_clusters = len([i for i in self.cm.sum_col.values() if i != 0])
if (n_classes == n_clusters == 1) or (n_classes == n_clusters == 0):
return 1.0
mutual_info_score = metrics.MutualInfo(self.cm).get()
entropy_true = _entropy(cm=self.cm, y_true=True)
entropy_pred = _entropy(cm=self.cm, y_true=False)
normalizer = self._generalized_average(entropy_true, entropy_pred)
normalizer = max(normalizer, np.finfo("float64").eps)
return mutual_info_score / normalizer
def __init__(self, cm: "metrics.ConfusionMatrix" = None):
self.cm = metrics.ConfusionMatrix() if cm is None else cm
self.pair_cm = metrics.PairConfusionMatrix(self.cm)
self.matthews_corr = metrics.MatthewsCorrCoef(cm=self.cm)
self.completeness = metrics.Completeness(cm=self.cm)
self.homogeneity = metrics.Homogeneity(cm=self.cm)
self.vbeta = metrics.VBeta(cm=self.cm)
self.q0 = metrics.Q0(cm=self.cm)
self.q2 = metrics.Q2(cm=self.cm)
self.pt = metrics.PrevalenceThreshold(cm=self.cm)
self.mutual_info = metrics.MutualInfo(cm=self.cm)
self.normalized_mutual_info = metrics.NormalizedMutualInfo(cm=self.cm)
self.adjusted_mutual_info = metrics.AdjustedMutualInfo(cm=self.cm)
self.rand = metrics.Rand(cm=self.cm)
self.adjusted_rand = metrics.AdjustedRand(cm=self.cm)
self.variation_info = metrics.VariationInfo(cm=self.cm)
(metrics.MicroF1(),
partial(sk_metrics.f1_score, average="micro", zero_division=0)),
(
metrics.WeightedF1(),
partial(sk_metrics.f1_score, average="weighted", zero_division=0),
),
(metrics.MCC(), sk_metrics.matthews_corrcoef),
(metrics.MAE(), sk_metrics.mean_absolute_error),
(metrics.MSE(), sk_metrics.mean_squared_error),
(metrics.Homogeneity(), sk_metrics.homogeneity_score),
(metrics.Completeness(), sk_metrics.completeness_score),
(metrics.VBeta(beta=0.5), partial(sk_metrics.v_measure_score, beta=0.5)),
(metrics.FowlkesMallows(), sk_metrics.fowlkes_mallows_score),
(metrics.Rand(), sk_metrics.rand_score),
(metrics.AdjustedRand(), sk_metrics.adjusted_rand_score),
(metrics.MutualInfo(), sk_metrics.mutual_info_score),
(
metrics.NormalizedMutualInfo(average_method="min"),
partial(sk_metrics.normalized_mutual_info_score, average_method="min"),
),
(
metrics.NormalizedMutualInfo(average_method="max"),
partial(sk_metrics.normalized_mutual_info_score, average_method="max"),
),
(
metrics.NormalizedMutualInfo(average_method="arithmetic"),
partial(sk_metrics.normalized_mutual_info_score,
average_method="arithmetic"),
),
(
metrics.NormalizedMutualInfo(average_method="geometric"),