The invention provides a
social network friend recommendation method fusing a trust degree, and relates to user similarity,
confidence factor calculation and fusion. The
social network topology-basedrecommendation focuses on known friends and ignores potential interested friends; the interest-based recommendation focuses on recommendation of strange users, thereby difficultly getting trust of users; and both the
social network topology-based recommendation and the interest-based recommendation do not consider behaviors of the users in a social network, thereby greatly influencing the accuracy, reliability and comprehensiveness of a recommendation result. The invention provides a recommendation method comprehensively considering social
network topology, user interests and social behaviors.Firstly, social similarity is calculated out according to common neighbors in the social network of the users, interest similarity is calculated according to keywords, and linear combination is performed. The social topology and the social behaviors of the users are comprehensively considered, a relationship confidence degree and a behavior confidence degree are calculated out, and fusion is performed to form a
confidence factor. Finally, the similarity and the
confidence factor are fused, so that the trust
degree of similarity calculation is improved, and a Top-N recommendation
list is generated.