0
$\begingroup$

Is there a way to perform clustering based on inertia threshold where each cluster inertia can’t exceed an inertia a specific inertia

I have tried hierarchical clustering with complete linkage with distance threshold, yet i can’t control the total distances of a cluster.

More info:

Im using precomputed distance matrix The main goal here is to divide destinations between drivers using roads network (distance matrix of destinations)

$\endgroup$

1 Answer 1

0
$\begingroup$

K-means objective function is to minimize total variance among all clusters. Hierarchial clustering objective function is to keep distance between two clusters within a threshold. Your objective function is to keep variance within clusters within a threshold. This does not fit into K-means or Hierarchial.

You need to use a mix of Kmeans and Hierarchial.

What you can do is run a loop from 2 till n_points. Set number of clusters as variable. Run K-means and estimate variance in all clusters. Stop loop when variance in any cluster goes above threshold.

Warn: This is computationally expensive exercise. You can try running not from 2 but a higher number

$\endgroup$
1

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.