An approach to reduce cold start in compound recommender systems using semantic technology and social networks

Today, the effort to build a "recommender system" with low precision and high speed in all conditions has become one of the most popular fields of research. Due to high percentage error, a basic method to build such systems is not usually being applied. In this research, two methods have been suggested in order to improve recommendations in recommender systems. The first suggested approach is a user-base method which predicts rates using similarity detection between target user and its neighbors. The second proposed approach is an item-base method which uses similarity detection between items, in order to predict possible rates of target user. Finally, the results show that combining semantic technology with social networks has reduced issues such as "cold start" and generally "data sparsity" in recommender systems.