In view of the high utilization rate of the social media nowadays, most people regard it as one of the channels to express their feelings, at the same time, have generated a large amount of useful information available for researchers to extract the sentiment and try to predict future values. However, most literature does not sufficiently focus on the pre-processing, sentiment analysis and post-processing of sentiment analysis. This study proposed a procedure for using a social media as data source and extracting the process of predicting the future values of sentiment. It is found that data pre-processing is complicated but cannot be avoided. This study proposed a procedure to reduce the scope of research and to facilitate the experimental comparison as a basis for the implementation.