NannyML / The-Little-Book-of-ML-Metrics

The book every data scientist needs on their desk.

https://www.nannyml.com/metrics

757 stars 60 forks source link

davies-bouldin-first-page #112

Open keswani-Rohitkumar opened 1 month ago

keswani-Rohitkumar commented 1 month ago

davis-bouldin first page

keswani-Rohitkumar commented 1 month ago

46 WIP

santiviquez commented 1 month ago

Looking pretty good so far 🔥

keswani-Rohitkumar commented 4 weeks ago

Thanks @santiviquez ! Happy for any recommendations / feedback on the above.

santiviquez commented 3 weeks ago

Hi @keswani-Rohitkumar, thanks for the great work!

I modified the metric's explanation and left an inline comment about When to use Davies Bouldin?.

I also think the formula can be improved. Currently, the colors are in $i$ and $j$, but I think it would make more sense if we color and add arrows to:

$k$ : total number of clusters.
$R_{ij}$: similarity between clusters $i$ and $j$
$s_i$: within-cluster scatter
$d_{ij}$: distance between the centroids

What do you think?