Outliers

Simple outlier detection by comparing distances between examples.

Signals

Inputs:

• Examples (ExampleTable)

  Attribute-valued data set.

• Distance matrix

  A matrix of example distances.

Outputs:

• Outliers (ExampleTable)

  Attribute-valued data set containing only examples that are outliers. Meta attribute Z-score is added.

• Inliers (ExampleTable)

  Attribute-valued data set containing only examples that are not outliers. Meta attribute Z-score is added.

• Examples with Z-scores (ExampleTable)

  Attribute-valued data set containing examples from input data with corresponding Z-scores as meta attribute.

Description

Outliers widget first computes distances between each pair of examples in input Examples. Average distance between example to its nearest examples is valued by a Z-score. Z-scores higher than zero denote an example that is more distant to other examples than average. Input can also be a distance matrix: in this case precalculated distances are used.

Two parameters for Z-score calculation can be choosen: distance metrics and number of nearest examples to which example’s average distance is computed. Also, minimum Z-score to consider an example as outlier can be set. Note, that higher the example’s Z-score, more distant is the example from other examples.

Changes are applied automatically.

Examples

Below is a simple example how to use this widget. The input is fed directly from the File widget, and the output Examples with Z-score to the Data Table widget.