1
$\begingroup$

I've just started out trying on ML. In WEKA, when I try to visualize a data set I find it hard to tell the class ratio for a certain nominal attribute value due to the differing amount of instances, which varies the staple height.

Example diagram

For instance, in the diagram above, I find it hard to compare the ratios (percent red and percent blue) between staple a and b.

Is there any way of adding percentages to the diagrams, or any other easy way to tell the ratios?

$\endgroup$
2
$\begingroup$

Questions about specific programs or coding (such as WEKA) are off topic here, but there is also a point of what visualization to use. Your question is an example of why stacked bar plots are not that great (see the work of William S. Cleveland for more on this).

It's not clear what each bar represents or what you are trying to show, but you could consider either unstacked bar plots (where the red and blue bars would be next to each other rather than stacked) or a mosaic plot. Or maybe some other graphic.

$\endgroup$
1
$\begingroup$

Following on from @Peter Flom's answer, here are some token plots, including also a direct graph of the percent red. Simply, if the percent red is a useful summary, or even the main concern, show it directly.

I guessed the numbers from your totals and eyeballing bar lengths. Clearly the principle is the main point.

I didn't use (have never used) WEKA and in any case agree with Peter on that detail.

enter image description here

enter image description here

$\endgroup$

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.