Change search
ReferencesLink to record
Permanent link

Direct link
Tweet Sentiment,
Sentiment Trend,
and a Comparison with
Financial Trend Indicators.
Norwegian University of Science and Technology, Faculty of Information Technology, Mathematics and Electrical Engineering, Department of Computer and Information Science.
2014 (English)MasteroppgaveStudent thesis
Abstract [en]

Background: As Twitter has become a global microblogging site, it’s influ- ence in the stock market has become noticeable. This makes tweets an interest- ing medium for gathering sentiment. A sentiment that might influence trends in the stock market. Motivation: If Twitter can be used to predict future prices in the stock mar- ket the casual investor would gain an advantage over the day-trader and the modern trading algorithms. Another interesting aspect is the role of Twitter in sentiment analysis. And how Twitters role as a data source influences trends in the stock market. Data and Experiments: Twitter is used as the data source. It provides easy access, lots of data, and many possibilities to use available metadata. To find the sentiment of a tweet we use two methods, counting positive and negative words(bag of words), and classifiers (SVM and Naive Bayes). We use moving average(MA) and average directional index(ADX) as trend indicators. We cal- culate MA and ADX with data from Oslo stock exchange, and we created our own indicators, based on MA and ADX, using data from Twitter. Then we compare the graphs. Findings: We explore the usage of lists of words, dictionaries, in sentiment analysis. And we look at data retrieval from Twitter and the trend we can create from it. To a varying degree we get positive results with the dictionaries, while the trend aggregation lacks the finesse and results we hoped for. Conclusion: Sentiment classification of tweets worked with both bag of words, and trained classifiers. We also managed to aggregate a trend based on senti- ment, but we found no correlation between the financial trend indicators and the sentiment indicators.

Place, publisher, year, edition, pages
Institutt for datateknikk og informasjonsvitenskap , 2014. , 106 p.
URN: urn:nbn:no:ntnu:diva-27032Local ID: ntnudaim:10013OAI: diva2:756639
Available from: 2014-10-17 Created: 2014-10-17 Last updated: 2014-10-17Bibliographically approved

Open Access in DiVA

fulltext(1194 kB)942 downloads
File information
File name FULLTEXT01.pdfFile size 1194 kBChecksum SHA-512
Type fulltextMimetype application/pdf
cover(184 kB)7 downloads
File information
File name COVER01.pdfFile size 184 kBChecksum SHA-512
Type coverMimetype application/pdf

By organisation
Department of Computer and Information Science

Search outside of DiVA

GoogleGoogle Scholar
Total: 942 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

Total: 277 hits
ReferencesLink to record
Permanent link

Direct link