WHICH ONE HAS BETTER PERFORMANCE IN EXPLAINING INDEXES’ RETURN USING SENTIMENT ANALYSIS? SOCIAL MEDIA VS ONLINE NEWS’ SENTIMENTS

• Main Author: Yoswara, Yogi
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/72463
• Institution: Institut Teknologi Bandung

The purpose of this study is specifically comparing the relation between social media versus online news’ sentiments on the return of sector indexes in Indonesia. This study compiles 1,975 data of public’s conversations in social media and 2,000 data of financial articles provided in online news about several companies or stocks during 3 months (January – March 2015). The main methodologies used in this study are sentiment analysis and linear regression. Sentiment analysis is conducted by developing keywords in Bahasa and inputting them to Semantria for Excel software to identify the sentiment score of each data, whereas linear regression is conducted by using SPSS software to identify the relation between sentiments and return of sector indexes. The focuses on this study are only two sector indexes in Indonesia: financial sector (JKFINA) and infrastructure, utility, and transportation sector (JKINFA). The contribution made in this study is the data of each company are pooled into one sector, so the return of sector indexes will be the outcome variable. The results of this study are (1) sentiments of social media and online news have positive relationship on the return of sector indexes, (2) the relationship of social media’s sentiments is slightly higher than online news’ sentiments, but they are approximately similar, and (3) more interestingly, if both sentiments (social media and online news) are combined, they generate stronger relationship. Next research can develop keywords in English and try to analyze the other financial outcome variable.