Opened 7 years ago

Last modified 4 years ago

#1304 new feature request

Implement spots view for regression and classification results

Reported by: swinkler Owned by: mkommend
Priority: low Milestone: HeuristicLab 3.3.x Backlog
Component: Problems.DataAnalysis Version:
Keywords: Cc:

Description (last modified by swinkler)

For regression results without temporal context and especially classification results a spots view (with a light gray lines between original and estimated values) would be nice.

Attachments (1)

ClassificationSpotsViewHL25.gif (9.6 KB) - added by swinkler 7 years ago.

Download all attachments as: .zip

Change History (8)

comment:1 Changed 7 years ago by swinkler

  • Summary changed from For visualizing regression results a spots view would be a good alternative to line charts to Implement spots view for regression and classification results

comment:2 Changed 7 years ago by gkronber

What is the added value of the spots view in comparison with the default classification view that is already implemented?

comment:3 Changed 7 years ago by gkronber

  • Milestone changed from HeuristicLab 3.3.3 to HeuristicLab x.x.x

comment:4 Changed 7 years ago by swinkler

  • Description modified (diff)

Changed 7 years ago by swinkler

comment:5 Changed 7 years ago by swinkler

The added value could be that such a spots view would enable users to zoom into certain ranges of samples and see original vs. estimated values. I have attached an example how this could look like. The view that is implemented now also shows a graphical representation of classifications, but such a spots view could give a quicker and more intuitive visual overview of the quality of the classifier. Of course, we can also decide that we do not need something like this.

comment:6 Changed 7 years ago by gkronber

I do not see the added benefit of zooming. However, I like that this chart shows misclassified examples as grey lines crossing the boundaries. The default classification view does not show the estimated and the actual class value at the same time.

I think that the view could be improved by sorting the examples by estimated value or something related to emphasize the misclassified examples and examples close to the separation threshold.

If the examples are simply ordered by the index in the dataset the chart is cluttered and there is no strong visual cue the information that should be transported by the chart. However, when the long grey lines are grouped the darker area spanned by the long grey lines visual indicates the number of misclassifications and the extent of estimation errors.

Last edited 7 years ago by gkronber (previous) (diff)

comment:7 Changed 4 years ago by gkronber

  • Version 3.3.2 deleted
Note: See TracTickets for help on using tickets.