Nowadays, Recommendation Systems (RS) play an important role in the eCommerce business and they have been proposed to exploit the potential of social networks by filtering information and offering useful recommendations to customers. As the personalization service is built to present the users with highly relevant set of items, the customer loyalty of the web companies can be improved. Collaborative Filtering (CF) is believed to be a suitable underlying technique for recommendation systems based on social networks, since it harvests information both from similar products and from peer users to infer a suggested item out of many for a user. Meanwhile, social networks provide the needed collaborative social environment. CF and its variants have been studied extensively in the literature on online recommender, marketing and advertising. However, most of the works were based on Web 1.0 where all the information necessary for the computation is by default assumed to be always available, as if it were readily stored in a database. In the distributed environment of Web 2.0 such as social networks, the required information by CF may either be incomplete or scattered over different sources. This poses new computational changes for Web 2.0 recommender. The system we proposed here is the Multi-Collaborative Filtering Trust Network Recommendation System, which combined multiple sources, by using MovieLens, Delicious and Facebook datasets, measured trust, temporal relation and similarity factors. After series of experiments, we found that the performance of recommendation system with considering above four aspects is much better than considering any other single/combined aspects.