2.7. Novelty and Outlier Detection¶
Many applications require being able to decide whether a new observation belongs to the same distribution as existing observations (it is an inlier), or should be considered as different (it is an outlier). Often, this ability is used to clean real data sets. Two important distinction must be made:
novelty detection:  

The training data is not polluted by outliers, and we are interested in detecting anomalies in new observations.  
outlier detection:  
The training data contains outliers, and we need to fit the central mode of the training data, ignoring the deviant observations. 
The scikitlearn project provides a set of machine learning tools that can be used both for novelty or outliers detection. This strategy is implemented with objects learning in an unsupervised way from the data:
estimator.fit(X_train)
new observations can then be sorted as inliers or outliers with a predict method:
estimator.predict(X_test)
Inliers are labeled 1, while outliers are labeled 1.
2.7.1. Novelty Detection¶
Consider a data set of observations from the same distribution described by features. Consider now that we add one more observation to that data set. Is the new observation so different from the others that we can doubt it is regular? (i.e. does it come from the same distribution?) Or on the contrary, is it so similar to the other that we cannot distinguish it from the original observations? This is the question addressed by the novelty detection tools and methods.
In general, it is about to learn a rough, close frontier delimiting the contour of the initial observations distribution, plotted in embedding dimensional space. Then, if further observations lay within the frontierdelimited subspace, they are considered as coming from the same population than the initial observations. Otherwise, if they lay outside the frontier, we can say that they are abnormal with a given confidence in our assessment.
The OneClass SVM has been introduced by Schölkopf et al. for that purpose
and implemented in the Support Vector Machines module in the
svm.OneClassSVM
object. It requires the choice of a
kernel and a scalar parameter to define a frontier. The RBF kernel is
usually chosen although there exists no exact formula or algorithm to
set its bandwidth parameter. This is the default in the scikitlearn
implementation. The parameter, also known as the margin of
the OneClass SVM, corresponds to the probability of finding a new,
but regular, observation outside the frontier.
References:
 Estimating the support of a highdimensional distribution Schölkopf, Bernhard, et al. Neural computation 13.7 (2001): 14431471.
Examples:
 See Oneclass SVM with nonlinear kernel (RBF) for visualizing the
frontier learned around some data by a
svm.OneClassSVM
object.
2.7.2. Outlier Detection¶
Outlier detection is similar to novelty detection in the sense that the goal is to separate a core of regular observations from some polluting ones, called “outliers”. Yet, in the case of outlier detection, we don’t have a clean data set representing the population of regular observations that can be used to train any tool.
2.7.2.1. Fitting an elliptic envelope¶
One common way of performing outlier detection is to assume that the regular data come from a known distribution (e.g. data are Gaussian distributed). From this assumption, we generally try to define the “shape” of the data, and can define outlying observations as observations which stand far enough from the fit shape.
The scikitlearn provides an object
covariance.EllipticEnvelope
that fits a robust covariance
estimate to the data, and thus fits an ellipse to the central data
points, ignoring points outside the central mode.
For instance, assuming that the inlier data are Gaussian distributed, it will estimate the inlier location and covariance in a robust way (i.e. whithout being influenced by outliers). The Mahalanobis distances obtained from this estimate is used to derive a measure of outlyingness. This strategy is illustrated below.
Examples:
 See Robust covariance estimation and Mahalanobis distances relevance for
an illustration of the difference between using a standard
(
covariance.EmpiricalCovariance
) or a robust estimate (covariance.MinCovDet
) of location and covariance to assess the degree of outlyingness of an observation.
References:
[RD1999]  Rousseeuw, P.J., Van Driessen, K. “A fast algorithm for the minimum covariance determinant estimator” Technometrics 41(3), 212 (1999) 
2.7.2.2. Oneclass SVM versus elliptic envelope¶
Strictlyspeaking, the Oneclass SVM is not an outlierdetection method, but a noveltydetection method: its training set should not be contaminated by outliers as it may fit them. That said, outlier detection in highdimension, or without any assumptions on the distribution of the inlying data is very challenging, and a Oneclass SVM gives useful results in these situations.
The examples below illustrate how the performance of the
covariance.EllipticEnvelope
degrades as the data is less and
less unimodal. svm.OneClassSVM
works better on data with
multiple modes.
For a inlier mode wellcentered and elliptic, the
svm.OneClassSVM is not able to benefit from the
rotational symmetry of the inlier population. In addition, it
fits a bit the outliers present in the training set. On the
opposite, the decision rule based on fitting an
covariance.EllipticEnvelope learns an ellipse, which
fits well the inlier distribution. 

As the inlier distribution becomes bimodal, the
covariance.EllipticEnvelope does not fit well the
inliers. However, we can see that the svm.OneClassSVM
tends to overfit: because it has not model of inliers, it
interprets a region where, by chance some outliers are
clustered, as inliers. 

If the inlier distribution is strongly non Gaussian, the
svm.OneClassSVM is able to recover a reasonable
approximation, whereas the covariance.EllipticEnvelope
completely fails. 
Examples:
 See Outlier detection with several methods. for a
comparison of the
svm.OneClassSVM
(tuned to perform like an outlier detection method) and a covariancebased outlier detection withcovariance.MinCovDet
.