2
$\begingroup$

I have a dataset of <user, item> pairs where each entry records which user bought which item. e.g.

<u1, i1>
<u1, i4>
<u2, i2>
<u3, i2>...

I created an encoded dataset with

no_of_features = no_of_users + no_of_items

and have set output variable y to 1, as each of those entries represented the user having bought that particular item.

Note: All y values are 1 in this case.

The encoded dataset looks like this:

user1 user2 user3 .... item1 item2 item3 item4 .... y

  1     0    0    ....   1     0     0     0   .... 1
  1     0    0    ....   0     0     0     1   .... 1
  0     1    0    ....   0     1     0     0   .... 1
  0     0    1    ....   0     1     0     0   .... 1

Now, I would like to know how to use fastFM to generate recommendations for cases <x, y> for any user x and any item y?

Clearly regression is out of the picture here. So should I use the classification or ranking approach of fastFM? Also how? e.g. If I use classification, do I need to generate instances with y=0? If I use ranking approach, do I arrange items by the output of ranking approach and recommend those with higher values?

$\endgroup$
1
  • 1
    $\begingroup$ Is there a better way of moving this question to datascience than duplicating it? For those reading, here it is on stackoverflow ;-) $\endgroup$ – Bartłomiej Twardowski Apr 18 '16 at 9:56

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.