Outliers

Sources & Credits:

• A Brief Overview of Outlier Detection Techniques

Overview

Let's give some clear definitions of outlier:

• Extreme values that deviate from other observations on data.

• An observation that diverges from an overall pattern on a sample.

Types

Depending on the number of dimensions that it affects:

• Univariate outliers: can be found when looking at a distribution of values in a single feature space.

• Multivariate outliers: can be found in a n-dimensional space.

Depending on the environment:

• Point outliers are single data points that lay far from the rest of the distribution.

• Contextual outliers can be noise in data, such as punctuation symbols when realizing text analysis or background noise signal when doing speech recognition.

• Collective outliers can be subsets of novelties in data such as a signal that may indicate the discovery of new phenomena.