How to make a recommendation algorithm?

Question

Algorithms

There is a table consisting of two fields:
1) an impersonal unique user identifier - session_id
2) product name (film, book, etc.)
Let's say one user logged in and two records were created in this table:
1) kvn
2) what where when
Then another user came in and two more entries were created: 1 ) what
where when
2) who wants to become a millionaire
How can I make an algorithm if more users come in, then output to them: "who wants to become a millionaire" (in that order) 2) Query: "who wants to become a millionaire" We issue: "what where when", "kvn" 3) Query: "what where When"
We issue: "kvn", "who wants to become a millionaire" or "who wants to become a millionaire", "kvn" (i.e. for equal related products, the order can be any)
I read the article https://habrahabr.ru/post/ 176549/ , but the formula there is too complicated and I did not understand it. How can I make a SQL query out of it?

Answer 1

No SQL, but you can play around with MVA in Sphinxsearch

Answer 2

but the formula there is too complicated and I did not understand it

The recommendation algorithm is one of the most algorithms )) Therefore, do not even count on simple solutions. Join:
Analysis of learning algorithms for collaborative recommender systems
engjournal.ru/articles/816/816.pdf