[Recommendation system practice] cold start problem

User cold start

1. Using user registration information

Coarse-grained demographic information for popular recommendations

2. Ask users to rate some items

The selected items need to have a degree of discrimination, xx algorithm

 

item cold start

3. Based on item content

Item similarity calculated by keywords, usually worse than cf, depending on the quality of the keywords mined

LDA topics are similar

 

Guess you like

Origin http://43.154.161.224:23101/article/api/json?id=324612965&siteId=291194637