liveBook · Manning

Sparse region

As explained in Outlier Detection in Python

An area in a two-dimensional plane containing relatively few observations compared with surrounding areas.

A sparse region is an area of a two-dimensional plane containing relatively few observations compared with surrounding areas. In outlier detection, records unlike most others tend to occupy such regions, making them useful signals for finding candidate outliers. section 10.1.2

Use sparse-region reasoning to discover feature subspaces in several ways:

  • Search domain-informed feature combinations whose values could be noteworthy. section 10.1.2
  • Prioritize associated features, because exceptions to their usual relationships can produce meaningful outliers. section 10.1.2
  • Use random subspaces when exhaustive searches for associations or sparse regions are too expensive; accept that selection can be suboptimal. section 10.1.2
  • Use exhaustive searches for small feature sets when interpretability matters, or use features tied to already known outliers for targeted detection. section 10.1.2

How Outlier Detection in Python covers it

How deep it goes
Explains the detection rationale and selection strategies but defers detailed algorithmic treatment to later detectors.
Examples
Uses staff-expense records and random, correlation-based, and baseball-statistic subspace listings as concrete feature-combination examples.
What you can do after
Evaluate feature subspaces by looking for large nearly empty regions associated with potential outliers.
How it connects
Connects sparse regions with outlier detection, feature subspaces, feature associations, random subspaces, and exhaustive search.
Cover of Outlier Detection in Python
liveBookmanning.com
Read Outlier Detection in Python on liveBook Get the book at manning.com

This explainer was generated by Manning's knowledge graph from the text of Outlier Detection in Python. The section links point to the parts of the book it draws on.