Totara Learn Open Discussions

Recommendation Engine Requirements

 
Harry Dixon
Recommendation Engine Requirements
بواسطة Monday, 17 May 2021, 7:21 AM - Harry Dixon
مجموعة Good learner 2023مجموعة Partnersمجموعة V18 Program Complete

Hi, 

We've been trying to use the 'Recommended For You' block on a dashboard to try and recommend some content. However, when I add this, no recommendations are currently returned. 

Are there some requirements for this to start returning results to a user? Such as the amount of data the user must have against their profile?

Let me know if any more information would be helpful. The block in the attached file has the default block settings. 

Thanks

Harry




لقد تمت إزالة مشاركة المنتدى هذه
Monday, 17 May 2021, 3:59 PM
تمت إزالة محتوى هذه المشاركة في المنتدى ولم يعد من الممكن الوصول إليها
Craig Eves
Re: Recommendation Engine Requirements
بواسطة Monday, 17 May 2021, 4:09 PM - Craig Eves (Totara Support)
مجموعة Totara

Hi Harry

The recommendations engine does need some data and time to process to provide recommendations.

I would expect some recommendation data if the recommendation engine has been set up and turned on for a few days.

Can you check under Plugins > Machine learning settings > Manage machine learning that the Recommendation engine is set to On?

Regards



Harry Dixon
Re: Recommendation Engine Requirements
بواسطة Tuesday, 18 May 2021, 3:25 AM - Harry Dixon
مجموعة Good learner 2023مجموعة Partnersمجموعة V18 Program Complete

Hi both, 


Thanks for the info. I can confirm that we've followed the configuration steps and that the recommendation engine is set to 'on'. It has been configured for at least a month or so, although it's a test site and doesn't have much real data. 


When we try to run the cron task directly for the recommendation engine, we get this error:

The data loading and processing/transformation of tenant 0 took  0 minutes and  0.02 seconds. The number of users or items is too small to run the recommendation engine. Skipping tenant 0.

So it does look to be an issue with the amount of data we have. Do you know how many users / items I'd need to add to expect this to start running? And what is meant by 'items' in this context? 


Thanks,

Harry

? ?
Re: Recommendation Engine Requirements
بواسطة Tuesday, 18 May 2021, 2:07 PM - ? ?
 

Hi Harry,

The message "The data loading and processing/transformation of tenant 0 took 0 minutes and 0.02 seconds. The number of users or items is too small to run the recommendation engine. Skipping tenant 0." does suggest that the recommendation engine is configured fine. The amount of data it sees, however, is not enough to run the ML algorithm.

The items in this context is the recommendable content (courses, workspaces, articles, and playlists).

By default, the recommendation engine will start generating the recommendations for users when there are at least 10 users and at least 10 items in the system.

Thanks,

Amjad

Harry Dixon
Re: Recommendation Engine Requirements
بواسطة Thursday, 20 May 2021, 8:27 AM - Harry Dixon
مجموعة Good learner 2023مجموعة Partnersمجموعة V18 Program Complete

Hi Amjad, 

Thanks for explaining. 

I've added more than 10 users and workspaces to our Totara instance. However, we still don't seem to be seeing any results for the Recommendation block. 

Is there any additional data you'd suggest adding? Or anything else I can try? 

Thanks

Harry

? ?
Re: Recommendation Engine Requirements
بواسطة Thursday, 20 May 2021, 3:12 PM - ? ?
 

Hi Harry

Thanks for getting back on this.

I should have explained this in my previous post. The RECOMMENDED FOR YOU block by default is set to show only the "Trending" content (popular content) which in your case may be empty being the fresh site.

To change the default RECOMMENDED FOR YOU block settings, click on the cog symbol in the top right corner of the block (as shown in the attached image)

Then click on "Configure Recommended for you block", and on the resultant page expand the "Custom block settings". From here, you can choose what type of "recommended content", you would like to see in the RECOMMENDED FOR YOU block. In your example, I would select "Workspaces" to show in this RECOMMENDED FOR YOU block (as in the attached image)



Please also note that the recommendations for users in case when there is very little data (as in your case) may not be as useful. For example, if you decide to show 10 Workspaces in one of the RECOMMENDED FOR YOU block when there were just 10 workspaces in total, each user will be recommended with the list of entire 10 Workspaces but ordered according to their taste (gathered from the users past interactions with the content, users profile data and the content data). So a good strategy here might be selecting fewer items to display (in the block settings) and extend this number as the site grows.

I hope this helps. Looking forward to hear back how it went for you.

Regards,

Amjad

Harry Dixon
Re: Recommendation Engine Requirements
بواسطة Tuesday, 25 May 2021, 1:00 AM - Harry Dixon
مجموعة Good learner 2023مجموعة Partnersمجموعة V18 Program Complete

Hi Amjad, 

Thanks again for the additional info. I noticed the additional config options when testing this before, and tried the 'workspaces' option. However, it doesn't seem to have worked as expected.

I've attached a screenshot of my current config for the recommendation block, in case you can see anything wrong.

So in summary, I can confirm:

  • The recommendation engine is configured
  • I've added at least 10 workspaces, and 10 users
  • I've set the 'Recommended for you' block to display workspaces


Is there anything else you'd suggest trying here?

Thanks

Harry



Valerii Kuznetsov
Re: Recommendation Engine Requirements
بواسطة Tuesday, 25 May 2021, 2:04 PM - Valerii Kuznetsov
مجموعة Totara

Hi Harry, 

Could you please create HD ticket for this? It will be easier to help you from there.

We have troubleshooting guide made in a step-by-step form which allows better understanding what part may be at fault: https://forms.gle/FrwPALsLAw3tAnvS7

Follow this form and let us know when it is completed, we will look in the reply and will be able to advise you on next steps (if the problem wouldn't be resolved by following it).