search

ankane / disco

Recommendations for Ruby and Rails using collaborative filtering
MIT License
565 stars 11 forks source link

Combining multiple "values" with weights #2

Closed jesster2k10 closed 4 years ago

jesster2k10 commented 4 years ago

Is it possible to use the implicit feedback recommendation system, where you specify a "value" but that value is based off of a number of different values, with weights.

It's still unclear to me how the "value" option works, in terms of the ranking. Does a higher value mean a user "likes" this item more or what's the correlation.

In my particular scenario, I'm trying to do:

  1. Recommend product categories to users
  2. Recommend users to users

And what I considered to be my inputs for the values are:

If I wanted to combine the values for the three parameters I noted above, with particular weights, say on the purchase_count how exactly would this be possible/what suggestions would you make?

jesster2k10 commented 4 years ago

I was looking at some other gems, and I suppose the best example to what I was trying to explain was the predictor gem https://github.com/Pathgather/predictor in the way you set up an input-matrix with the assigned weights

ankane commented 4 years ago

Hey @jesster2k10, a higher value indicates a stronger preference (since we don't get any negative feedback, we won't know if they actually liked it). I'm not familiar with the best way to combine values (typically you want an algorithm to learn the weights), but you can try manually combining them to see how that goes (value = page_views + 10 * likes + 100 * purchase_count).

If you find a better way, please let me know!