• Post Reply Bookmark Topic Watch Topic
  • New Topic
programming forums Java Mobile Certification Databases Caching Books Engineering Micro Controllers OS Languages Paradigms IDEs Build Tools Frameworks Application Servers Open Source This Site Careers Other Pie Elite all forums
this forum made possible by our volunteer staff, including ...
Marshals:
  • Campbell Ritchie
  • Ron McLeod
  • Paul Clapham
  • Jeanne Boyarsky
  • Liutauras Vilda
Sheriffs:
  • Rob Spoor
  • Bear Bibeault
  • Tim Cooke
Saloon Keepers:
  • Tim Moores
  • Stephan van Hulst
  • Tim Holloway
  • Carey Brown
  • Piet Souris
Bartenders:
  • Frits Walraven
  • Himai Minh

Build an ML Recommender System

 
Greenhorn
Posts: 5
  • Mark post as helpful
  • send pies
    Number of slices to send:
    Optional 'thank-you' note:
  • Quote
  • Report post to moderator
Hi,

Do you have any suggestions on how to deal with the cold start problem (i.e. that you miss substantial data in the beginning the recommendation process)?

Thanks!
Laura
 
Greenhorn
Posts: 9
  • Mark post as helpful
  • send pies
    Number of slices to send:
    Optional 'thank-you' note:
  • Quote
  • Report post to moderator
Hi,

When a new user or item is introduced into the system, it is called cold start. I would claim that you can't solve it because you can't personalise without knowing anything (no data). If you look at the research, cold start problems are usually solved by getting data from the side (which you could then claim is not cold start any longer). Or, if the user is subscribing to a new streaming service, the user will often be asked to select a list of content the user likes, and then the recommendations can be build using those.

But if you have a cold start situation, you can use popularity recommendations, i.e. non-personalised recommendations, trying out the content most people like. You can also look into diversifying the recommendations such that there is a greater chance that you show something the user would like.

Cold start items can be handled in several ways, the most simple is to add the new items into the recommendations at random or use content-based similarities to figure out what the new item is similar to. Or if we are talking streaming then simple do the new arrivals section and users will quickly create data.

Have a look at chapter 6 of my book Practical Recommender Systems if you want to know more.
reply
    Bookmark Topic Watch Topic
  • New Topic