Abstarct: Nowadays social media have received more attention. Public and private opinions about a wide variety of subjects are expressed by users continually via numerous social media. Twitter is one of the social media that is gaining popularity in recent decade. Twitter offers organizations a fast and effective way to analyze customer’s perspectives toward the critical to success in the market place. Sentiment analysis or opinion mining is the process of extracting people’s opinions, perspectives and sentiments about a particular topic. Sentiment analysis in the domain of micro-blogging is a new research topic so it has still the potential to research and develop in this field. Lots of works has been done on sentiment analysis of user reviews, documents and articles. Analysis on the above items differ from twitter mainly because of the limit of 280 characters per tweet which forces the user to express his/her opinion compressed and short. The best results reached in sentiment classification use machine learning techniques such as Naive Bayes and Support Vector Machines. In this research we proposed a method for sentiment analysis in social networks. In this respect, we have tried to improve Naïve Bayes classification by focusing on preprocessing and feature selection steps. The proposed method has the detection accuracy of 83.37 percent. In addition, results shows that using n-gram features provide the best results and combining it with attribute weighting baxsed on gain ratio obtains a remarkable improvement in the results.