Abstract: Recently so many users who are different in power/interest trade news on social networking sites like Arabic Twitter and share their views on current affairs. These opinions/comments can't be used to make good decisions or boost output in a particular area without differentiating users according to their closeness/interest to this area. The main objective of this paper is to present a generic automatic model to analyze user opinions to make valuable decisions in a particular area based on the degree of user closeness/interest to this area. The proposed model combines rough set theory, Mendelow's power-interest model, and data mining decision-making techniques. Rough set theory based on Mendelow's power-interest model supports the identification and classification of users by their account features. Unsupervised k-means would then be used to cluster their replies/opinions into positive, negative, or neutral. The result generated from the classification of users and the clustering phase of Arabic replies/opinions supports the making of valuable/important decisions in a particular area. A case study is carried out to demonstrate the effectiveness and accuracy of the proposed model.