Y-oHr-N / kenchi

A scikit-learn compatible library for anomaly detection

kenchi

This is a scikit-learn compatible library for anomaly detection.

Dependencies

• Required dependencies

  1. numpy>=1.13.3 (BSD 3-Clause License)
  2. scikit-learn>=0.20.0 (BSD 3-Clause License)
  3. scipy>=0.19.1 (BSD 3-Clause License)

• Optional dependencies

  1. matplotlib>=2.1.2 (PSF-based License)
  2. networkx>=2.2 (BSD 3-Clause License)

Installation

You can install via pip

pip install kenchi

or conda.

conda install -c y_ohr_n kenchi

Algorithms

• Outlier detection

  1. FastABOD [8]
  2. LOF [2] (scikit-learn wrapper)
  3. KNN [1], [12]
  4. OneTimeSampling [14]
  5. HBOS [5]

• Novelty detection

  1. OCSVM [13] (scikit-learn wrapper)
  2. MiniBatchKMeans
  3. IForest [10] (scikit-learn wrapper)
  4. PCA
  5. GMM (scikit-learn wrapper)
  6. KDE [11] (scikit-learn wrapper)
  7. SparseStructureLearning [6]

Examples

import matplotlib.pyplot as plt
import numpy as np
from kenchi.datasets import load_pima
from kenchi.outlier_detection import *
from kenchi.pipeline import make_pipeline
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler

np.random.seed(0)

scaler = StandardScaler()

detectors = [
    FastABOD(novelty=True, n_jobs=-1), OCSVM(),
    MiniBatchKMeans(), LOF(novelty=True, n_jobs=-1),
    KNN(novelty=True, n_jobs=-1), IForest(n_jobs=-1),
    PCA(), KDE()
]

# Load the Pima Indians diabetes dataset.
X, y = load_pima(return_X_y=True)
X_train, X_test, _, y_test = train_test_split(X, y)

# Get the current Axes instance
ax = plt.gca()

for det in detectors:
    # Fit the model according to the given training data
    pipeline = make_pipeline(scaler, det).fit(X_train)

    # Plot the Receiver Operating Characteristic (ROC) curve
    pipeline.plot_roc_curve(X_test, y_test, ax=ax)

# Display the figure
plt.show()

References

[1]	Angiulli, F., and Pizzuti, C., "Fast outlier detection in high dimensional spaces," In Proceedings of PKDD, pp. 15-27, 2002.

[2]	Breunig, M. M., Kriegel, H.-P., Ng, R. T., and Sander, J., "LOF: identifying density-based local outliers," In Proceedings of SIGMOD, pp. 93-104, 2000.

[3]	Dua, D., and Karra Taniskidou, E., "UCI Machine Learning Repository," 2017.

[4]	Goix, N., "How to evaluate the quality of unsupervised anomaly detection algorithms?" In ICML Anomaly Detection Workshop, 2016.

[5]	Goldstein, M., and Dengel, A., "Histogram-based outlier score (HBOS): A fast unsupervised anomaly detection algorithm," KI: Poster and Demo Track, pp. 59-63, 2012.

[6]	Ide, T., Lozano, C., Abe, N., and Liu, Y., "Proximity-based anomaly detection using sparse structure learning," In Proceedings of SDM, pp. 97-108, 2009.

[7]	Kriegel, H.-P., Kroger, P., Schubert, E., and Zimek, A., "Interpreting and unifying outlier scores," In Proceedings of SDM, pp. 13-24, 2011.

[8]	Kriegel, H.-P., Schubert, M., and Zimek, A., "Angle-based outlier detection in high-dimensional data," In Proceedings of SIGKDD, pp. 444-452, 2008.

[9]	Lee, W. S, and Liu, B., "Learning with positive and unlabeled examples using weighted Logistic Regression," In Proceedings of ICML, pp. 448-455, 2003.

[10]	Liu, F. T., Ting, K. M., and Zhou, Z.-H., "Isolation forest," In Proceedings of ICDM, pp. 413-422, 2008.

[11]	Parzen, E., "On estimation of a probability density function and mode," Ann. Math. Statist., 33(3), pp. 1065-1076, 1962.

[12]	Ramaswamy, S., Rastogi, R., and Shim, K., "Efficient algorithms for mining outliers from large data sets," In Proceedings of SIGMOD, pp. 427-438, 2000.

[13]	Scholkopf, B., Platt, J. C., Shawe-Taylor, J. C., Smola, A. J., and Williamson, R. C., "Estimating the Support of a High-Dimensional Distribution," Neural Computation, 13(7), pp. 1443-1471, 2001.

[14]	Sugiyama, M., and Borgwardt, K., "Rapid distance-based outlier detection via sampling," Advances in NIPS, pp. 467-475, 2013.