Hi Orange3 team,
I'm using the CN2 Rule Induction model in Orange3 and really enjoying its ability to generate human-readable rules.
However, I’d like to avoid rules that contain "not equals" (≠) conditions. In my use case, such rules are less actionable, and I’d prefer the model to explore alternative rules using "equals" (=) conditions, even if they cover different subsets of the data.
Unfortunately, filtering out ≠ rules after the fact isn’t a good solution because many times those are the only rules returned, and removing them leaves me with nothing. What I’d really like is a way to steer the model to prefer or restrict itself to = conditions during rule induction, ideally via a configuration option or parameter.
Is this currently possible in Orange3? If not, would you consider supporting it?
Thanks so much for your work and for considering this feature!
Hi Orange3 team,
I'm using the CN2 Rule Induction model in Orange3 and really enjoying its ability to generate human-readable rules.
However, I’d like to avoid rules that contain "not equals" (≠) conditions. In my use case, such rules are less actionable, and I’d prefer the model to explore alternative rules using "equals" (=) conditions, even if they cover different subsets of the data.
Unfortunately, filtering out ≠ rules after the fact isn’t a good solution because many times those are the only rules returned, and removing them leaves me with nothing. What I’d really like is a way to steer the model to prefer or restrict itself to = conditions during rule induction, ideally via a configuration option or parameter.
Is this currently possible in Orange3? If not, would you consider supporting it?
Thanks so much for your work and for considering this feature!