[jira] [Commented] (MATH-1516) Define an interface for ranking a list of clusters

["Chen Tao (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/MATH-1516?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17054157#comment-17054157 ] 

Chen Tao commented on MATH-1516:
--------------------------------

There are many clusters evaluation algorithm:
[scikit-learn clustering-performance-evaluation|https://scikit-learn.org/stable/modules/clustering.html#clustering-performance-evaluation]
They can be divided into 2 categories: “External Measures” and "Internal Measures".
The function signatureis can be decided by the category the evaluation algorithm belong to.

Althought the score is the higher the better for most of these evaluation algorithm, but there is a special case:
[Davies-Bouldin Index|https://scikit-learn.org/stable/modules/clustering.html#davies-bouldin-index]

There also some simplified evaluation like SumOfClusterVariances, the score is the lower the better.

If there is a training application program, replaceable evaluator is necessary, the evaluator algorithm has the responsibility to isolate the rank rule. This should be considered in the design.

> Define an interface for ranking a list of clusters
> --------------------------------------------------
>
>                 Key: MATH-1516
>                 URL: https://issues.apache.org/jira/browse/MATH-1516
>             Project: Commons Math
>          Issue Type: Sub-task
>            Reporter: Gilles Sadowski
>            Assignee: Gilles Sadowski
>            Priority: Minor
>             Fix For: 4.0
>
>
> [On the "dev" ML|https://markmail.org/message/z4qr3fcsg5emt2nn] it has been suggested to create a functional interface for unequivocally defining the quality of a clustering:
> * a valid ranking must be positive,
> * better clustering is conveyed through higher ranking.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)