kth.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • 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
ChatGPTs prediktion av nyhetsartiklars inverkan på aktier: Kan det slå index?
KTH, School of Electrical Engineering and Computer Science (EECS).
KTH, School of Electrical Engineering and Computer Science (EECS).
2024 (Swedish)Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE creditsStudent thesisAlternative title
ChatGPTs prediction of news articles affect on stocks : Can it beat index? (Swedish)
Abstract [en]

As the global economy digitizes and data becomes a central resource, the need for tools that can interpret large amounts of data increases. Advances in machine learning enable more sophisticated analysis that mimics human interpretation. One particularly interesting application is sentiment analysis in financial markets. With the introduction of Large Language Models (LLMs) such as ChatGPT, we can now analyze financial news with a scale and precision that was previously impossible. This work investigates whether LLMs can predict stock movements based on company-specific news articles and whether these analyses can be used in a trading strategy to outperform market indices. By analyzing news articles from the top 20 US news sources for Nasdaq-100 companies for one month in 2024, the performance of ChatGPT 3.5 and ChatGPT 4 is compared. The results show that the models, especially with the long strategy, can outperform the index, especially in market capitalization-weighted investments. The study also highlights challenges such as efficient market theory, where new technologies are rapidly adopted, reducing the opportunities for outperformance. The results indicate that while LLMs can improve predictions and trading strategies, further research is needed to address their limitations and maximize their potential.

Abstract [sv]

I takt med att den globala ekonomin digitaliseras och data blir en central resurs ökar behovet av verktyg som kan tolka stora mängder data. Framsteg inom maskininlärning möjliggör mer sofistikerad analys som efterliknar mänsklig tolkning. En särskilt intressant tillämpning är sentimentanalys på finansmarknaderna. Med introduktionen av stora språkmodeller (LLM) som ChatGPT kan vi nu analysera finansiella nyheter med en skala och precision som tidigare var omöjlig. I detta arbete undersöks om LLM:er kan förutsäga aktierörelser baserat på företagsspecifika nyhetsartiklar och om dessa analyser kan användas i en handelsstrategi för att överträffa marknadsindex. Genom att analysera nyhetsartiklar från de 20 största amerikanska nyhetskällorna för Nasdaq-100-företag under en månad 2024 jämförs prestandan hos ChatGPT 3.5 och ChatGPT 4. Resultaten visar att modellerna, särskilt med den långa strategin, kan överträffa index, särskilt i marknadsvärdesviktade investeringar. Studien belyser också utmaningar såsom teorin om effektiva marknader, där ny teknik snabbt tas i bruk, vilket minskar möjligheterna till överavkastning. Resultaten visar att LLM kan förbättra förutsägelser och handelsstrategier, men att det krävs ytterligare forskning för att ta itu med deras begränsningar och maximera deras potential.

Place, publisher, year, edition, pages
2024. , p. 8
Series
TRITA-EECS-EX ; 2024:415
Keywords [en]
LLM, ChatGPT, Stock Prediction, News articles, AI
National Category
Computer and Information Sciences
Identifiers
URN: urn:nbn:se:kth:diva-351222OAI: oai:DiVA.org:kth-351222DiVA, id: diva2:1886724
Supervisors
Examiners
Available from: 2024-09-19 Created: 2024-08-03 Last updated: 2024-09-19Bibliographically approved

Open Access in DiVA

fulltext(1082 kB)124 downloads
File information
File name FULLTEXT01.pdfFile size 1082 kBChecksum SHA-512
164fc8462f497fb96f4bff43793aa28cb1049d5f93afb60ad75265d128440d60af3a3a998a76a645b94907f7118983e28a973e35fc44b4ca50b4cf4995a060de
Type fulltextMimetype application/pdf

By organisation
School of Electrical Engineering and Computer Science (EECS)
Computer and Information Sciences

Search outside of DiVA

GoogleGoogle Scholar
Total: 124 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

urn-nbn

Altmetric score

urn-nbn
Total: 260 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • 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