The marketplace market is developing independently of the platform on which online stores operate. The need for the formation of a full range of services is still unmet, especially when it comes to the choice of medicines or cosmetics. AI-powered recommender systems should solve the main problems faced by numerous sites. How this should be done can be considered on the subject of shops offering all kinds of creams, lotions, cosmetics and skin care products.
For such cases, the principle of collaborative filtering is well suited, which builds predictions based on already known preferences, and gives recommendations for yet unknown preferences of completely different users. The principle is simple - once a given assessment of a phenomenon or product, left earlier, is the basis for similar assessments of other phenomena and products in the future. The advantage of collaborative filtering is its individual "sharpening" for each client, despite the fact that the information rationale for the forecast is collected from the responses of thousands of other people.
This approach uses three methods for creating recommender systems. The first is collaborative filtering, the second is content-based recommendations, and the third is hybrid.
The whole recommendation system looks like this.
– . , :
, ;
() ;
, .
, , , , , , , . . . , , . , .
– . , . . , , . . . , :
. , . , . . .
?
, , . , , , , , .
, . «» , , .
. , «». «» «» . , «» - , . «»- .
, , , .
: 15 15 , 14 . 210 (1514) , «» . . – , . :
Xi = WXi
, .
, :
N – N , mm – , dd — , XijXij — i- j- . .
, , .
–
, , - -. . , .. , .
, , «» , , . , , «» . , , «» . , , , . , , .
. , - , . . , « » , .
, , – IP-, . . , . , . , .