Recommendation list returned is shorter than pageSize

[Sang Tran]

Hi,

I get recommendations list from a trained "recommended_for_you" model.

I have more than 700 products in product catalog, and set "pageSize": 100 in the request.
But in I only have 6 products in returned recommendation list.

Could you give me any hints to what causes this behavior or is there any way can I expand the list ?

[Peng Ren]

Hi Sang,

This could be working as intended. "pageSize" only indicates the maximum results to be returned but not guarantee to return the requested size every time.

Recommendations AI will try best to return items to match the page size; however, there are many factors could cause the gap. Could you please check the following things in your data/request:

1. Do you have a large set of products covered in your user event data? It is important to have good coverage of product catalog in your user logs since training is based on use event data. For example, if you have 700 products in your catalog, but only 100 of them can found in user events, then model can only return those in the events.
2. Do you have lots of out of stock products? We filter out out of stock items in the results for you.
3. Did you use filter tags in your prediction request? Overusing of filter tags could lead to a small number of nomination candidates.

Best,

Peng

[Sang Tran]

Thanks Peng for your quick response.

I have checked as your response:

1. + 2. :

    I have more than 900 products imported into catalog of Rec.AI, in which more than 700 products with availability="IN_STOCK".

    I have counted user events in 5 recently days (from July 03 to July 08). There are nearly 90% products shown in these events.

    And I inputed these events as "detail-page-view" events into Rec.AI

3. I use console UI to get recommendation list as below picture, so I don't think the list is affected by tag filtering:

I really dont understand why the list is shortened when there are scores also returned in it, and users can cut off according the list as the need based on these scores

Thanks & Regards.

Sáng Trần.

[Peng Ren]

Hi Sang,

Could you email me your project number for us to further debug for you? Thanks!

Best,

Peng

--
You received this message because you are subscribed to a topic in the Google Groups "cloud-recommendations-users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/cloud-recommendations-users/yPJuCOUKCrU/unsubscribe.
To unsubscribe from this group and all its topics, send an email to cloud-recommendatio...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cloud-recommendations-users/df0775a5-19b7-4216-b3b6-bd62d7976cefn%40googlegroups.com.

[Peng Ren]

Hi Sang,

Based on our internal team debugging, it seems that your "Recommended for you" model has the diversification turned on by default; however, you only have very limited categories(only two categories from our finding) in your product catalog. Therefore, with diversification on, we only return limited results from each categories and that's why you see very few prediction results in total. You can probably revamp your product catalog to have more categories or turn off the diversification on your model. Thanks!

Best,

Peng

[Sang Tran]

Dear Peng,

Thanks for your explanation.

My team is going to build a model "recommended_for_you" without diversification.

Further, from my point of view, there should be a document or an example to explain how the option "diversification" works.
It's hard for users (for example my case) to guess what is affecting the recommendation list.

Thank you for your great support, Peng !

Thanks & Regards.

Sáng Trần.