Abstract
This paper investigates the design and evaluation of a personalized Recommender System (RS) using implicit social trust from Online Social Networks (OSNs), particularly to solve new users' recommendation problems. The proposed system builds implicit trust based on the interrelation between an active user and his/her friends in the popular social microblogger Twitter, by considering aspects such as retweet actions and followers/followings lists. The measured trust values are used to vote for friends' opinions held in the posted tweets about a certain product such as movies. The higher trust parameters to a friend the more his/her opinions anticipate in recommendations encounter. Firstly, Friends' opinions are obtained by a probabilistic sentiment analysis technique to extract the opinions in form of multi-point scale of ratings from short tweets. Secondly, trust relation aspects are extracted from user's friends accounts. Further, a genetic algorithm is used to optimize social trust parameters. Thirdly, this paper considers the Support Vector Regression algorithm (SVR) to predict ratings for the active user. Our experimental results show that the proposed approach outperforms several related works in terms of accuracy using real world data from Twitter. These results can have a promising effect when solving new users, so called cold-start problem, to the systems by integrating users' OSNs.