Label color in model explanation

[Sjunja]

What's wrong?

Look at the picture. The legend is from blue to ted. Where those yellows, oranges ang greens come from?

How can we reproduce the problem?

What's your environment?

• Operating system: Windows 10
• Orange version: 3.32
• How you installed Orange: Standalone for windows

[PrimozGodec]

Thank you for reporting the issue. I transferred it to the orange3-explain repository (home for the explaining addon).

In the case of categorical variables (for models which do not one-hot encode them), we plot category colors instead of whether the value is high or low since it is impossible to plot that for categorical variables. However, the main issue here is that there is just a legend for numerical variables (high low), but the user does not know what colors are for categorical variables. We should rethink how to address the issue.

[Sjunja]

Thank you! I think that this is an issue of specific models (Naïve Bayes). The left is the Naïve Bayes, the right is Logistic regression (and Boosting, and Random forest, and Neural Network) on the same data: Kind regards, Timur Syunyakov, MD, PhD Senior researcher, Learning Center, SBFH «Clinical Psychiatric Hospital №1 named after Alekseev» Leading expert (Coordinator on Statistical Data Analysis), International Centre for Education and Research in Neuropsychiatry (ICERN), Samara State Medical University, Samara, Russia Scientific Adviser, Republic Special Educational and Research Narcology Centre, Tashkent, Uzbekistan