Which criteria to use while Evaluating minimum number of cluster before k-means ?

Hi Matlabers,

During evaluating mimimum number of clusters of specific dataset, how to choos the criteria among following ?

The function name is "evalclusters". We have to choose one critera from following _

1.DaviesBouldin

2. CalinskiHarabasz

3.gap

How to decide one citeria for the evaluation ?

Regards,

Koustubh

Answer 1

[Jalaj Gambhir]

Hi,

What criteria would work best for you totally depends on the data you are using. You need to try different criteria values and/or read more about how they are working. You can read more about the methods here.

1. Gap criteria
2. Davies-Bouldin Criterion
3. Calinski-Harabasz Criterion