1
$\begingroup$

I have gathered a large amount of qualitative data and am now looking to cluster it so as to make sense of it. For this, I am using Biolab's Orange.

In my data, specific values may co-occur in a given feature, or they may not. I am wondering how I could cluster the data (either in Orange or other software), where values that co-occur would be seen as two values, rather than one string.

To make matters clearer, imagine I have a feature X, with the possible values A and B. The values can occur in the following way: A, B, A and B.

Question: how can I cluster my data without "A and B" being treated as a separate string, but rather "A" and "B" co-occurring?

$\endgroup$
0
$\begingroup$

It would appear that you have a data processing task. You could use two columns 'A Occurrence' and 'B Occurrence'. If that value occurs for the record that index would contain a 1. If that value does not occur for that record the index would contain a 0.

Example:

X | A Occurrence | B Occurrence
A | 1 | 0
A and B | 1 | 1

$\endgroup$
0
$\begingroup$

There are plenty of well-established methods for this.

Read up on Jaccard index. Actually I don't like the current Wikipedia article much, because I consider the computer vision example quite stupid, for example. I think the discussion should rather be based on the original biological species use case.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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