umu.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard1
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Portfolio investment strategy based on Twitter sentiment
Umeå University, Faculty of Science and Technology, Department of Mathematics and Mathematical Statistics.
2017 (English)Independent thesis Advanced level (professional degree), 20 credits / 30 HE creditsStudent thesis
Abstract [en]

This paper investigates if it is possible to create a portfolio investment strategy by looking at the sentiment (i.e. are they positive or negative) of twitter data for ten companies, five IT companies and five fashion companies. 764 340 tweets were collected during the study which spanned 60 trading days, and of those tweets, 483 946 where from the IT companies and the rest from the fashion companies. The tweets were collected in a Python program using Twitters API, and then analyzed and classified in another Python program using three different Naive Bayes classifiers that had been trained on a training set consisting of positive and negative text. The sentiment results were then used to create two different portfolios where one was based solely on sentiment and the other one was a combination of sentiment and market capitalization, the ratio used was determined by testing. Those portfolios were then compared against a market capitalization portfolio and a Sharpe portfolio.

I found that for the IT companies the portfolio based solely on sentiment performed decently, but was the worst of the four portfolios. The combination portfolio performed well and when comparing it to the Sharpe portfolio and the market capitalization portfolio, it might even be the preferable strategy depending on the investor’s appetite for risk as it had the highest ratio between return and standard deviation. For the fashion companies the sentiment portfolio performed very poorly. The combination portfolio performed decently, but that was only because it consisted mainly (85%) of the market capitalization portfolio which performed the best of all strategies and thereby “saving” the combination portfolio. The poor performance of the sentiment portfolio for the fashion companies might in part be explained by the fact that there were almost twice as many tweets for the IT companies, making the sentiment less accurate and less reliable for the fashion companies when compared to sentiment of the IT companies. It might also be that there is more irrelevant stuff being tweeted about when it comes to the fashion companies, causing the sentiment portfolio to performworse.

Place, publisher, year, edition, pages
2017.
National Category
Mathematics
Identifiers
URN: urn:nbn:se:umu:diva-136679OAI: oai:DiVA.org:umu-136679DiVA: diva2:1113099
Educational program
Master of Science in Engineering and Management
Supervisors
Examiners
Available from: 2017-06-21 Created: 2017-06-21 Last updated: 2017-06-21Bibliographically approved

Open Access in DiVA

fulltext(1148 kB)91 downloads
File information
File name FULLTEXT01.pdfFile size 1148 kBChecksum SHA-512
8a360750f31e0854df7cbe625162d41f3d2fed8513c43ade1adf41c9cf835cc60b83fbc9389322482c895c3d85a5184ea6fdf63437089cae30b1aaea35653100
Type fulltextMimetype application/pdf

By organisation
Department of Mathematics and Mathematical Statistics
Mathematics

Search outside of DiVA

GoogleGoogle Scholar
Total: 91 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: 56 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard1
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf