Contact: Fares Meghdouri - [email protected]

Observers-based Data Modeling - paper: "Modeling Data with Observers" (pending)

pyodm can be installed using pip by running pip install git+https://github.com/CN-TU/pyodm

Note that in order for ODM to work with an M-Tree core, the implementation (package) in M-Trees needs to be installed. The repository is private and will be available soon.

Please note that many parameters can be adjusted in order to build a representative model refer to the paper for more information.

import pyodm

# create a new model with default parameters
model = pyodm.ODM(random_state=1)

import numpy
#import pandas

# read the data
X = np.load('my_dataset.npy')

#or
#X = pandas.read_csv('my_dataset.csv').values

# model the data
model.fit(X)

# access the array of observers
print(model.observers)

# access the array of radius
print(model.radius)

# access the array of populations
print(model.population)

In order to get the outlierness score of a set of points (based on an ODM model), run the foolowing after fitting a model

# read the data
X_test = np.load('my_test_dataset.npy')

# get the outlierness scores
outlierness_scores = model.outlierness(X_test)

One can convert the outlierness score into a binary label (outlier/inlier) using the following

# read the data
X_test = np.load('my_test_dataset.npy')

# convert the outlierness scores into binary labels using a contamination threshold
predictions = model.predict(X_test) 

to get the label of the closest observer to a set of points use

# read the data
X_test = np.load('my_test_dataset.npy')

# return the predicted label of each test point (refering to `model.observers`)
predicted_labels = model.labels(X_test)

# return a dictionnary of the current parameters
model.get_params()

This will return a dictionnary of parameters used to build the model.\

Example1: Three datasets in which datapoints are represented in gray and the ODM model in red each with a different configuration.

Example2: Convergence path of an observer.

Example3: Two clusters datasets with two observers.

Example4: Five clusters datasets with six observers.