MDO
MDO (Mahanalobis detection outliers) is a method based on the inference of some parameters (means vertors and precisions matrice) of gaussian mixture with the EM algorithm to define mahanalobis distance and a scoring.
Installation
Use the package manager pip to install MDO.
pip install MDO
Usage
from MDO.main import MDO
params = {
"inference_type"='bayesian',
"n_components" : 10,
"covariance_type" : 'full',
...
} #Parameters for bayesian Gaussian mixture or the usual one
outliers_dectetion = MDO()
outliers_dectetion = outliers_dectetion.fit(X, **params) # Procede to inference for finding parameters (means and precision matrice)
List_scoring_global = MDO.get_scoring() # returns global scoring
List_scoring_local = MDO.get_scoring("local") # returns local scoring