liveBook · Manning

Probability theory

As explained in Outlier Detection in Python

A numerical quantity representing the degree of likelihood assigned to an event or class prediction.

Probability expresses how likely an event or class prediction is. In classifier-based outlier detection, it is often more useful than a binary label because it supports ranking and threshold decisions. section 4.3.3

Interpret classifier probabilities cautiously:

  • A probability may be poorly calibrated, so 0.89 does not necessarily mean an 89% true likelihood. section 4.3.3
  • Well-fitted probabilities can still rank cases: a record scored 0.89 is more likely to be an outlier than one scored 0.6. section 4.3.3
  • Classifier probabilities identify records resembling labeled outlier types, not every statistically unusual record. section 4.3.3

How Outlier Detection in Python covers it

How deep it goes
Explains calibration, ranking, and classifier limitations without deriving probability theory or statistical-outlier detection.
Examples
Compares classifier scores of 0.89 and 0.6 for two records.
What it assumes
Assumes classifier predictions and labeled outlier examples, while distinguishing specific outlier types from statistical anomalies.
What you can do after
Apply it via classifier-based outlier detection to rank cases while avoiding literal interpretations of uncalibrated scores.
How it connects
Connects probability with classifiers, outlier detection, binary labels, decision boundaries, and statistical outliers.
Cover of Outlier Detection in Python
liveBookmanning.com
Read Outlier Detection in Python on liveBook Get the book at manning.com

Also explained in

Compare how 16 books explain probability theory →

Related concepts

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.