liveBook · Manning

DML detector

As explained in Outlier Detection in Python

A DML detector is an outlier detector based on a DML modeling approach and capable of using internally trained classifiers.

A DML detector identifies outliers by modeling the distinction between real and synthetic records. It uses the leaf-node set reached by each prediction, rather than directly using a classifier’s outlierness score. section 11.5

The detector can use an internally trained classifier, commonly a Random Forest, and depends on synthetic data realistic enough to support meaningful discrimination. Classifier choice, tuning, and synthetic-data construction may require experimentation. section 11.5

Synthetic records may be produced with:

  • Left-to-right generation that preserves associations between features. section 11.5
  • Modified Gaussian mixture models, kernel density estimation, doped records, or simulations. section 11.5

How Outlier Detection in Python covers it

How deep it goes
Distinguishes DML from real-versus-fake scoring but defers detailed treatment while emphasizing the more commonly used alternative.
Examples
Uses left-to-right generation, modified Gaussian mixture models, kernel density estimation, doped records, and simulations to create synthetic data.
What you can do after
Apply via synthetic-data generation and classifier training, while tuning both components for the detection task.
How it connects
Connects DML with synthetic data, Random Forests, real-versus-fake outlier detection, generative models, and classifier training.
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.