liveBook · Manning

Total cohesion

As explained in Advanced Algorithms and Data Structures

Total cohesion is a measure of how closely points are grouped within their clusters, here based on cosine distance to a cluster center.

Total cohesion measures how closely points group around their cluster centers by using cosine distance. It is especially useful for categorical features. section 12.5

Use total cohesion as one way to assess clustering quality. It resembles distance to a cluster center, but replaces Euclidean distance with cosine distance; intra-cluster distance provides a different alternative based on average pairwise distance. section 12.5 section 12.6

How Advanced Algorithms and Data Structures covers it

How deep it goes
Defines the metric and contrasts it with Euclidean and pairwise-distance measures, without deriving its calculation.
Examples
Uses categorical features as the motivating case for preferring cosine-based total cohesion.
What it assumes
Assumes familiarity with clusters, cluster centers, and distance-based clustering evaluation.
What you can do after
Evaluate clustering quality with a cosine-distance measure suited to categorical features.
How it connects
Connects clustering with cluster centers, cosine distance, Euclidean distance, and intra-cluster distance.
Cover of Advanced Algorithms and Data Structures
liveBookmanning.com
Read Advanced Algorithms and Data Structures on liveBook Get the book at manning.com

This explainer was generated by Manning's knowledge graph from the text of Advanced Algorithms and Data Structures. The section links point to the parts of the book it draws on.