A big data approach to opinion analysis in social media

Abstract

Some recent studies have suggested that public opinions expressed in social media may be correlated with various social issues. To find out what actually can be discovered in social media data, we need data mining. Data mining approaches that can handle massive amounts of data have recently been referred to as big data approaches. In this thesis, we propose a big data approach to handling opinion analysis. By means of a literature review on works related to social media opinion mining, we found that the message content or the texts in social media are ambiguous and unstructured and are often in the form of short sentences. In other words, extracting clear and accurate opinions is difficult. To do so, in this study we propose an Ontology-based adapted social media data collection system called the OACM system, as well as a fuzzy big data algorithm called FMM. Through several sets of comparable experiments, the proposed OACM system and FMM have shown their effectiveness. The OACM can optimize system resource scheduling efficiently, and can effectively speed up the collection of large amounts of data in a relatively short time from multiple social media sources. Meanwhile, the FMM can identify opinions expressed in social media data more accurately when compared with other data mining algorithms, and can reduce computing complexity as well as processing time significantly.