Mix of continuous and categorical variables association rules

[PARODBE]

Hi,

Is there any option of use association rules when you have a mix of variables continuous and categorical? Like CBA algorithm... Thanks

[rasbt]

Good question, something like this is currently not implemented. For the current implementations, the items have to be discrete categories.

[PARODBE]

And do you think that It could be posible? Are there other libraries like CBA of arules in r or pyarc in python. For It IS of your interest

[NimaSarajpoor]

@PARODBE
I think there are some packages in R (e.g. see CBA). As explained in the package, one idea is to discretize the continuous variable to make it categorical. Then, you should(?) be able to use existing version of mlxtend for association rule discovery.

[PARODBE]

Thanks @NimaSarajpoor I know this option but sometimes when you discretize a continuous variable you could loss information, for this I prefer a methods which incorporates the option of working with continuous variables...

[rasbt]

Yes, discreticing the continuous variables as @NimaSarajpoor suggested would be the easiest. An option/algorithm that support continuous variables directly would of course be nice, but I currently wouldn't have the capacity to work on something like that in the foreseeable future, unfortunately.