Now a day the boom of social media is very popular to share their viewpoints to their friends is very easy by various social networking platforms. In this paper, we face information overloading problem. We mine the information from reviews what the user understood. Then user’s preferences and an accurate recommendation is done. It consists of some important factors purchasing records by the user, category of product and their location. We propose a sentiment based rating prediction. It improves the prediction accuracy in the recommender systems. At the first we propose a social user measurement approach sentimentally and then it calculates each user’s sentiments based on the products. At the second it not only views a user’s own sentiment attribute will consider the interpersonal sentiment influence. Then, it also considers product reputation sentimental distribution of a user can infer it reflects customer’s evaluation. Here we combine all the three factors where user sentiment that is similar, interpersonal sentimental influence and reputation of items it considers all into the recommender system and make accuracy in rating prediction. It makes compulsory for the user to pose their viewpoints as reviews before buying another product to know the quality of the product. It also considers the performance evaluation for all the three factors in the real word. As the result, it helps to improve the recommendation performance.