liveBook · Manning

Minimum support

As explained in Outlier Detection in Python

Minimum support is the lowest fraction of records in which an item set must occur to qualify as frequent.

Minimum support is the lowest fraction of rows in which an item set must occur to qualify as frequent. In Apriori-based outlier detection, it controls which recurring combinations are retained. section 3.2.4

Choose the threshold as a trade-off:

  • Raise it to avoid an excessive number of weak item sets, but excessively high values can leave even strong inliers with too few matches. section 3.2.4
  • Lower it to retain rarer combinations, but low values can produce meaningless item sets, reduce interpretability, slow execution, and make outliers resemble inliers. section 3.2.4

SpeedDating item sets

Using Apriori with minimum support 0.3 retains combinations occurring in at least 30% of rows. The resulting frequent item sets include ambitious ratings of [0–15] with support 0.797326 and a two-feature combination with support 0.593817.

How Outlier Detection in Python covers it

How deep it goes
Defines the row-fraction threshold and explains runtime, interpretability, and discrimination trade-offs, but does not derive Apriori internally.
Examples
Uses binned SpeedDating preference features and reports item sets retained at a 0.3 threshold.
What it assumes
Assumes tabular features can be one-hot encoded and teaches threshold effects through categorical or binned values.
What you can do after
Apply it via Apriori by setting min_support before generating frequent item sets.
How it connects
Connects frequent item sets, Apriori, one-hot encoding, and outlier detection.

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.