Differentially Private Recommendation System Based on Community Detection in Social Network Applications

Algorithm 2

CDAI recommendation method.

Input: ;

;

;

;

The known labels of users;

Output:The unknown labels of users;

Phase 1: Get the coarse-grained , List A

Get a new social network graph G’

Initiallize Each Node to a Separated Community

repeat

fordo

fordo

Remove i from its community, place to j’s community

Compute the composite modularity gain

end for

Choose j with maximun positive gain ( if exists) and move i to j’s community

Otherwise i stays in its community

end for

until No further improvement in modularity

Gets the label probability in each community,

Find , Assign to

Assign to

Get the coarse-grained

Phase 2: Get the fine-grained for the unknow labels of users in community. Get the list B.

fordo

ifthen

Users of the independent community, the labels remain the same

end if

end for

Get final