Message Boards Message Boards

Clustering techniques for a large multi-asset portfolio

Posted 9 years ago
Attachments:
POSTED BY: Igor Hlivka
5 Replies

enter image description here - Congratulations! This post is now a Staff Pick! Thank you for your wonderful contributions. Please, keep them coming!

POSTED BY: Moderation Team

Igor

Thank you for including the notebook as well. That way I can play around with the results.

Michael

POSTED BY: Michael Kelly
Posted 9 years ago

Hi Igor,

Thanks for the example. Crystal clear. The catch with qualitative data will be to transform it and/or to modify the clustering methods so that the clusters are meaningful. It will be interesting.

Gregory

POSTED BY: Gregory Lypny

Hello Gregory

Yes, clusters can be multi-dimensional. Below is a simple example using numerical data - in fact just 3 dimensional array to be able to visualise it. I am asking <FindClusters> command to generate 4 clusters

cdata = RandomVariate[NormalDistribution[0.01, 0.01], {100, 3}];
FindClusters[cdata, 4];
ListPlot3D[%]

This will look as follow: Clusters for 3D data

As you can see, you can work with multi-dimensional data, but visualisation of the clustered objects can be more difficult if you move beyond 3 dimensions. If the data is non-numeric, you are likely to work with strings. You can still try using the default distance function or select specific method such as DamerauLevenshteinDistance or HammingDistance.

To execute it: build your data as nested list where each sub list contains the data you want. Then simply apply the FindClusters command to your data object.

Hope this answers your query. Kind regards Igor

POSTED BY: Igor Hlivka
Posted 9 years ago

Hi Igor,

Thanks for sharing this. I'd like to learn how to construct multidimensional clusters, for example, observations comprised of the value of a company, its industry (Standard Industrial Classification), the state or city of its headquarters, and the names of its beneficial owners.

Gregory

POSTED BY: Gregory Lypny
Reply to this discussion
Community posts can be styled and formatted using the Markdown syntax.
Reply Preview
Attachments
Remove
or Discard

Group Abstract Group Abstract