摘要：Many peoples use Facebook to connect and share their views on various issues, with the majority of user-generated content consisting of textual information. Since there is so much actual data from people who are posting messages on their situation in real time thoughts on a range of subjects in everyday life, the collection and analysis of these data, which may well be helpful for political decision or public opinion monitoring, is a worthwhile research project. Therefore, in this paper doing to analyze for public text post on Facebook stream in real time through environment Hadoop ecosystem by using apache spark with NLTK python. The post or feeds are gathered form the Facebook API in real time the data stored database used Apache spark to quick query processing the text partitions in each data nodes (machine). Also used Amazon cloud based Hadoop cluster ecosystem into processing of huge data and eliminate on-site hardware, IT support, and other operational difficulties and installation configuration Hadoop such as Hadoop distribution file system and Apache spark. By using the principle of decision dictionary, emotion analysis is used as positive, negative, or neutral and execution two algorithms in machine learning (naive bias & support vector machine) to build model predict the outcome demonstrates a high level of precision in sentiment analysis.
