Financial sentiment analysis from big data
Heikura, Jarkko (2018)
Tässä tietueessa ei ole kokotekstiä saatavilla Treposta, ainoastaan metadata.
Heikura, Jarkko
2018
Kauppatieteiden tutkinto-ohjelma - Degree Programme in Business Studies
Johtamiskorkeakoulu - Faculty of Management
Hyväksymispäivämäärä
2018-05-09
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:uta-201805141670
https://urn.fi/URN:NBN:fi:uta-201805141670
Tiivistelmä
For decades research evidence supported the efficient market hypothesis, which assumed fully rational investors, random walk and informational efficiency. Until recently studies documented anomalies to efficient market hypothesis and the behavioral research offered explanations to these anomalies. More specifically attention, investor, and financial news related sentiment literature find over- and under-reaction patterns after positive or negative events. As well behavioral economic scholars suggest that people are making systematic errors, that does not cancel.
Today's major news agencies not only compete to break economic data, but they also sell millisecond-time-stamped breaking new, which has been available just in recent years. The vast amount of today's data can help economists to create meaningful statistics on economic behavior. In this thesis, psychological and behavioral elements in stock price determination have been studied with the help of big data source; digitally available financial news. More specific the target was to quantify news sentiments and explain stock price formation by combining big data analysis methods with classical econometric methods. The objective of the study is Finnish stock market (OMXH) from which stock market and Finnish economic news data were collected between 2004 and 2017.
It was found that sentiments cause abnormal returns in Finnish stock markets. Key statistical findings are: (1) positive firm-specific stories predict higher firm valuation in the next months; (2) negative firm-specific news predict under-reaction within a month; (3) firm-specific news sentiment, consumer confidence, and size of the firm are elements of market reaction.
The approach of using artificial intelligence together with behavioral theory revealed only part of its potential. In the future as more data comes available scholars can generate more accurate predictions and insides of human economic behavior.
Today's major news agencies not only compete to break economic data, but they also sell millisecond-time-stamped breaking new, which has been available just in recent years. The vast amount of today's data can help economists to create meaningful statistics on economic behavior. In this thesis, psychological and behavioral elements in stock price determination have been studied with the help of big data source; digitally available financial news. More specific the target was to quantify news sentiments and explain stock price formation by combining big data analysis methods with classical econometric methods. The objective of the study is Finnish stock market (OMXH) from which stock market and Finnish economic news data were collected between 2004 and 2017.
It was found that sentiments cause abnormal returns in Finnish stock markets. Key statistical findings are: (1) positive firm-specific stories predict higher firm valuation in the next months; (2) negative firm-specific news predict under-reaction within a month; (3) firm-specific news sentiment, consumer confidence, and size of the firm are elements of market reaction.
The approach of using artificial intelligence together with behavioral theory revealed only part of its potential. In the future as more data comes available scholars can generate more accurate predictions and insides of human economic behavior.