kenchi 0.7.2

A set of python modules for anomaly detection

kenchi

This is a set of python modules for anomaly detection.

Requirements

• Python (>=3.6)

• matplotlib (>=2.1.1)

• networkx (>=2.0)

• numpy (>=1.14.0)

• pandas (>=0.22.0)

• scikit-learn (>=0.19.1)

• scipy (>=1.0.0)

Installation

You can install via pip

pip install kenchi

or conda.

conda install -c y_ohr_n kenchi

Usage

import matplotlib.pyplot as plt
import numpy as np
from kenchi.datasets import make_blobs
from kenchi.outlier_detection import SparseStructureLearning

train_size       = 1000
test_size        = 250
n_outliers       = 10
n_features       = 25
centers          = np.zeros((1, n_features))

# Generate the training data
X_train, y_train = make_blobs(
    n_inliers    = train_size,
    n_outliers   = 0,
    n_features   = n_features,
    centers      = centers,
    random_state = 1
)

# Generate the test data that contains outliers
X_test, _        = make_blobs(
    n_inliers    = test_size - n_outliers,
    n_outliers   = n_outliers,
    n_features   = n_features,
    centers      = centers,
    random_state = 2,
    shuffle      = False
)

# Fit the model according to the given training data
det              = SparseStructureLearning().fit(X_train)

# Plot the anomaly score for each training sample
det.plot_anomaly_score(X_test, linestyle='', marker='.')

plt.show()

License

The MIT License (MIT)

Copyright (c) 2017 Kon