liveBook · Manning

Unclassified outlier

As explained in Outlier Detection in Python

An outlier that is not assigned to any recognized category of useful or known outliers.

An unclassified outlier is an outlier that does not resemble any recognized category of useful outlier. Novel outliers may otherwise be mislabeled as inliers or assigned to an existing category. section 15.3.2

Detect it by training one One-Class Support Vector Machine per known outlier category. Records inside a category’s learned boundary match that type; records outside every boundary remain unclassified. section 15.3.2

Category-specific OCSVM boundaries

For a dataset containing four known outlier types, train an OCSVM on examples of one type, such as a small cluster in the bottom right, then predict across the full dataset. Points matching none of the category boundaries are unclassified outliers. section 15.3.2

How Outlier Detection in Python covers it

How deep it goes
Explains category-specific training and decision boundaries, but does not develop model-selection or evaluation procedures.
Examples
Uses four known outlier clusters, including a small bottom-right cluster, and demonstrates an OCSVM implementation.
What it assumes
Assumes recognized useful-outlier categories and training examples for each category.
What you can do after
Apply via category-specific OCSVM models to identify records matching no known useful-outlier category.
How it connects
Connects outlier detection with classifiers, One-Class Support Vector Machines, and decision boundaries.
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.