search

biolab / orange3-associate

🍊 :package: Frequent itemsets and association rules mining for Orange 3.
Other
65 stars 37 forks source link

associate rule in Orange GUI only supports basket file? #24

Closed duohappy closed 6 years ago

duohappy commented 6 years ago
Associate version

1.15

Orange version

3.11

Expected behavior

After learning about Orange association-rules help document in http://orange3-associate.readthedocs.io/en/latest/scripting.html , I know that 0-1 file is not supported perfectly, because it counts 0 as 'existing' items. But the function of onehot transforms 0-1 to False-True, it can get right result in python script

I try to apply association rules in Orange GUI, which is better in displaying result. I save a csv file, such as, test.csv milk, bread, tomato, apple True, False, True, False False,True,False, False ........... but it counts False as 'existing' items, which is different from python script and makes me so confused. I think it maybe improved.

Reading Orange help document

Orange supports 4 data type, Continuous, Discrete, String, Date, so it can't recognize bool type such as False-True. If I want to get right result in Orange GUI, I should use basket file, which is not so convenient.

kernc commented 6 years ago

How would you have it improved? Consider that you know beforehand False (in True or False) stands for lack of something. How would Orange know? Categorical (discrete) variables can have arbitrary (number of) values, such as male/female, none/few/many, ...

duohappy commented 6 years ago

@kernc ,of course, you are right, and I agree with you. Maybe, improving Orange association rules is not necessary.

The reason why I open this issue is that I can not get right result in Orange GUI after only reading help document of association rules. In help file, True-False data like that

>>> X
array([[False,  True, ...,  True, False],
       [False,  True, ...,  True, False],
       [ True, False, ..., False, False],
       ...,
       [False,  True, ...,  True, False],
       [ True, False, ..., False, False],
       [ True, False, ..., False, False]], dtype=bool)

I can get right result in python script, which leads me to use True-False data in Orange GUI.

Adding some description like "In Orange GUI, data file should be basket file" in help file is easier to understand.