Trust-based matrix factorization, which extends SVD++ with trust information.
Trust information. The weights of the connections can be integer or float (binary relations can represented by 0 or 1).
The learning algorithm with the supplied parameters.
Trained recommender. Signal Predictor sends the output signal only if input Data is present.
Latent features of the users.
Latent features of the items.
Latent features of the implicit information.
Latent features of the trust information.
TrustSVD widget uses a biased regularized algorithm which makes use of implicit feedback information and trust information to factorize a matrix into four low rank matrices as it’s explained in Guibing Guo, Jie Zhang, Neil Yorke-Smith, TrustSVD: Collaborative Filtering with Both the Explicit and Implicit Influence of User Trust and of Item Ratings