Clustering with multiple distance vectors

[mvangisbergen]

Hi All, I would like to perform a clustering algorithm based on multiple distance vectors. These vectors describe the distances between the same entities but are based on different variables. As far as I’m aware there is no possibility to directly perform clustering based on multiple distance vectors; there can only be one distance vector. I see two possible work-arounds:

1. I compute a weighted average over the multiple distance vectors so I end up with a single distance vector and then I can perform k-mediods. I’m not sure whether it is possible to calculate with distance vectors in KNIME though.

2. I perform a hierarchical clustering based on the distance vectors and perform a clustering based on categorical variables (the clusters). For this I would need a clustering algorithm that support categorical variables (something like k-modes) but I’m not sure KNIME supports this.

Any advice on this would be much appreciated! Best, Maxim

[janina]

Hallo @mvangisbergen,

do you know about the Aggregated Distance node. Maybe this node would help for your use case.

Best,
Janina

[mvangisbergen]

Hi Janina, I had not heard about this node, and it is definitely what I was looking for… Many thanks! Best, Maxim