Sentiment Analysis on Twitter by Using Maximum Entropy and Support Vector Machine Method

Abstract

With the advancement of social media and its growth, there is a lot of data that can be presented for research in social mining. Twitter is a microblogging that can be used. In this event, a lot of companies used the data on Twitter to analyze the satisfaction of their customer about product quality. On the other hand, a lot of users use social media to express their daily emotions. The case can be developed into a research study that can be used both to improve product quality, as well as to analyze the opinion on certain events. The research is often called sentiment analysis or opinion mining. While The previous research does a particularly useful feature for sentiment analysis, but it is still a lack of performance. Furthermore, they used Support Vector Machine as a classification method. On the other hand, most researchers found another classification method, which is considered more efficient such as Maximum Entropy. So, this research used two types of a dataset, the general opinion data, and the airline's opinion data. For feature extraction, we employ four feature extraction, such as pragmatic, lexical-grams, pos-grams, and sentiment lexical. For the classification, we use both of Support Vector Machine and Maximum Entropy to find the best result. In the end, the best result is performed by Maximum Entropy with 85,8% accuracy on general opinion data, and 92,6% accuracy on airlines opinion data.

Comparison of Forensic Tools on Social Media Services Using the Digital Forensic Research Workshop Method (DFRWS)

“…Twitter is one of the social media in Indonesia with 19.5 million users [4]. Twitter social media provides services that allow its users to utilise it in their daily activities such as sharing whatever is around them quickly and easily [5]. Apart from being a place to seek entertainment or obtain information, social media also has the potential to become a place and even a medium for crime.…”

Sentiment Analysis of Public Responses on Indonesia Government Using Naïve Bayes and Support Vector Machine

“…a lot of data has been obtained that can be used as research material [1], [2]. One of the benefits that can be obtained from this data is being able to find out the opinions or sentiments of social network users regarding a topic that is being discussed on social media [3], [4]. Because from the results of sentiment analysis can be obtained new knowledge that can be used to make decisions [5].…”