Implementation of eALS

[DanielRauser]

Hello,

I really admire your work with this package and it helps me greatly for my bachelor thesis. I’m not that experienced with Recommender Systems and put in a lot of effort in understanding the last month. Still I have a few questions and I would really appreciate it, if someone could provide me with answers:

1. I‘m mainly working with implicit Data, which means that I want to use mainly ALS_implicit_real_Rating_Matrix. Can this matrix include Ratings something like that: (1) if customer viewed product, (2) if customer added product to cart, (3) if customer purchased product?
2. I’m wondering if I can implement the logic behind eALS. eALS weights non existent ratings by the item popularity which reduces the sparsity and therefore increases the performance. Is it possible to edit the RealRatingMatrix like that and can the underlying algorithm handle this data input or is it required that either eALS is implemented in the recommenderlab package or I have to build it myself or switch to Python?
3. There isn’t an option to tune parameters right? So I have to do this through recosystem and then implement it in parameter options right?

Like I already said, I really would appreciate an answer to my questions but I understand that there is probably bigger fish to fry :)

Have a great day,

Daniel

[mhahsler]

Hi Daniel,

The Code for the R implementation in recommenderlab is available here: https://github.com/mhahsler/recommenderlab/blob/master/R/RECOM_ALS.R

It has several parameters (see code) and I assume that adapting the weights should be possible. If you can suggest code changes to allow for different weights, then I would be happy to incorporate them into the package.

Btw, is there a paper for eASL?

-Michael

[DanielRauser]

Hello Michael,

thank you very much for your reply! I will have a look into it. Yes there are papers available:

• Fast Matrix Factorization for Online Recommendation with Implicit Feedback (He et. al.)
• Improving Implicit Recommender Systems with View Data (Ding et. al.)

Kind Regards,

Daniel