objectif

Détecter anomalies via forêts aléatoires.

code minimal

from sklearn.ensemble import IsolationForest
import numpy as np
X = np.r_[np.random.RandomState(0).normal(size=(100,2)), [[8,8]]]
clf = IsolationForest(random_state=0).fit(X)
pred = clf.predict([[8,8]])[0]
print(pred in (-1,1))

utilisation

from sklearn.svm import OneClassSVM
print(hasattr(OneClassSVM(kernel="rbf"), "fit"))

variante(s) utile(s)

from sklearn.covariance import EllipticEnvelope
print(hasattr(EllipticEnvelope(), "fit"))

notes

• Marque -1 pour anomalies; calibrer contamination si connue.