Capturing investor sentiment from Big Data: the effects of online social media on SET50 index

Abstract

This research aims to introduce a market sentiment index which can be used as a leading indicator for the Stock Exchange of Thailand (SET). This new index is constructed from Big Data extracted from online social media. The data used in this project are the daily firm-level and market data, market capitalizations of firms and the SET, the total trading value in the SET and the values of stocks bought and sold by investor types, the data on S&P global index, the implied volatility index of S&P500 (known as VIX) and the exchange rate of Thai Baht. Meanwhile, tonal words regarding 50 companies listed in the SET50 index are extracted from selected online/social media using search engines and an application programming interface (API). Note that 50 firms included in the analyses and index construction change every six months following the announcement of the SET. The daily data will be collected for the period 2015 - 2018. Machine Learning algorithms are employed to conduct a bag-of-word analysis which will count the number of positive, negative and neutral tonal words for 50 firms in a set of articles over a 4-year period. The findings show that the introduced sentiment index has significant relationships with the changes in SET50 index. The sentiment index can signal changes in SET50 index from day t to day t+3. The analysis results indicate that this sentiment index can act as a leading indicator of the SET50 index and thus the SET index.

More information

Identifiers

Catalogue record

Export record