[Question] - Sequence of transaction basket about seq2pat HOT 5 CLOSED

Great video -- thank you for sharing!

The short answer is no, not out of the box, since the library consider sequences with singleton items (as opposed to a tuple like (A1, A2, B) being an item).

Some data transformation might be possible during pre-processing to feed the input accordingly, OR, dropping the tuples to check if, in your example, A1, A2, A still shows up as frequent at the end.

That said, under the hood for the MDD propagation, there are great parallels with the PrefixSpan algorithm like the pseudo-projection etc.

from seq2pat.

The short answer is no, the software is independent of the definition of an item.

If the item of interest is a tuple as in <A1,A2,B>, then you can transform the data to map them, say to X. Likewise, <A,D> to represent another item Y. Then, the event sequence of tuples [["A1", "A2", "B"], ["A", "D"]] becomes [X, Y].

In essence, what we call "items" does not really matter (it can be single items, or tuples) and is not an off-the-shelf transformation within the library, we leave that definition to the specific application.

from seq2pat.

Actually I was interested in investigating the relation within each single tuple element, such as A1 -> D but I think this is not the expected usage.
Thank you for your answer and your work!

from seq2pat.

I see, makes sense. If there is a sequence relationship "within" the tuple, may be it is not a static tuple then? As in, items should remain as A1, A2, B and the sequence becomes A1 -> A2 -> B whereby you would have "two" sequences in your first row in the example [["A1", "A2", "B"], ["A", "D"]] --> ["A1", "A2", "B"] as one sequence and ["A", "D"] as another sequence. Just an idea.

Thanks for the positive feedback!

from seq2pat.

I guess yes, tuple size is definitely not static.
My question was trying to address whether this algorithm was fitting the use case of PrefixSpan, such as, in this case

["A1", "A2", "B"], ["A", "D"]

evaluating independently single element A1, A2 and B implying A or D and eventually subsets of tuples, e.g. {A1,A2} -> {A}
I attached here this video to explain better the scenario.

from seq2pat.