Topological Influence-Aware Recommendation on Social Networks

Algorithm 1

The proposed recommendation method RDISI with direct and indirect social influence.

Input: List of training triples (user id, item id, rating), list of tuples (user id, trustee id), the

dimensionality of user feature vector and item feature vector , the learning rate , the

parameters , , and .

Output: User-user trust matrix, user-item rating matrix, local influence vector , global

influence vector , user feature matrix , and item feature vector .

1: Generating user-item rating matrix and user-user trust matrix

for each triple   do

end for

for each tuple   do

end for

2: Calculating local influence and global influence

for   do

  calculate

  calculate   using Eq. (3)

  calculate

  calculate using Eq. (4)

end for

3: Initialize and randomly

while not convergence do

  calculate according to Eq. (8)

  calculate according to Eq. (9)

  update

  update

end while