kenchi 0.5.0

A set of python modules for anomaly detection

kenchi

This is a set of python modules for anomaly detection.

Requirements

• Python (>=3.5)

• matplotlib (>=2.0.2)

• numpy (>=1.11.2)

• pandas (>=0.20.3)

• scipy (>=0.18.1)

• scikit-learn (>=0.18.0)

Installation

You can install via pip.

pip install kenchi

Usage

import matplotlib.pyplot as plt
import numpy as np
from kenchi.outlier_detection import GaussianOutlierDetector

train_size   = 1000
test_size    = 100
n_outliers   = 10
n_features   = 10
rnd          = np.random.RandomState(0)
mean         = np.zeros(n_features)
cov          = np.eye(n_features)

# Generate the training data
X_train      = rnd.multivariate_normal(
    mean     = mean,
    cov      = cov,
    size     = train_size
)

# Generate the test data that contains outliers
X_test       = np.concatenate([
    rnd.multivariate_normal(
        mean = mean,
        cov  = cov,
        size = test_size - n_outliers
    ),
    rnd.uniform(
        low  = -10.0,
        high = 10.0,
        size = (n_outliers, n_features)
    )
])

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

# Plot anomaly scores for test samples
det.plot_anomaly_score(X_test)

plt.show()

License

The MIT License (MIT)

Copyright (c) 2017 Kon