Fast training and prediction: it is possible to train and predict with
a large number of detection models in PyOD by leveraging SUOD framework.
See SUOD Paper
and SUOD example.
frompyod.models.suodimportSUOD# initialized a group of outlier detectors for accelerationdetector_list=[LOF(n_neighbors=15),LOF(n_neighbors=20),LOF(n_neighbors=25),LOF(n_neighbors=35),COPOD(),IForest(n_estimators=100),IForest(n_estimators=200)]# decide the number of parallel process, and the combination method# then clf can be used as any outlier detection modelclf=SUOD(base_estimators=detector_list,n_jobs=2,combination='average',verbose=False)