dc.contributor.advisor | Sirnes, Espen | |
dc.contributor.author | Abdollahi, Hooman | |
dc.date.accessioned | 2024-10-24T10:33:34Z | |
dc.date.available | 2024-10-24T10:33:34Z | |
dc.date.embargoEndDate | 2026-11-07 | |
dc.date.issued | 2024-11-07 | |
dc.description.abstract | This dissertation stands at the intersection of finance and machine learning to empirically delve into the intricate relation between media sentiment and market volatility, covering various aspects of this dynamic. Through three research papers, the thesis presents a detailed investigation into how media—ranging from traditional news to social media platforms—shapes market behaviors and sentiments, influencing market volatility.
In the first paper, we employ a novel approach to quantify media sentiment’s impact on market volatility across various financial markets. Using advanced natural language processing techniques, we extract semantic sentiment from news headlines and social media posts. Our findings reveal a time-varying connection between media sentiment and market volatility, indicating how changes in media-driven sentiment can lead to fluctuations in market volatility.
In the second paper, we shift our focus to the predictive power of sentiment in financial markets. We examine the role of media sentiment, derived from both news and social media, in forecasting oil price volatility, given its critical role in the global economy. Employing a hybrid model that integrates sentiment analysis with forecasting models, we reveal the distinct predictive abilities of news versus social media sentiment. The analysis demonstrates that incorporating media sentiment significantly enhances the accuracy of these predictions.
In the third paper, we explore the information transmission pattern from real and fake news to market volatility. Through an advanced analysis of political news stories, we classify news items as either real or fake and then measure their respective impacts on stock market volatility. Our study uncovers the distinct patterns of volatility associated with each type of news, highlighting the more pronounced yet short-lived influence of fake news compared to the sustained but steadier impact of real news. | en_US |
dc.description.abstract | Denne avhandlingen befinner seg i skjæringspunktet mellom finans og maskinlæring for å empirisk utforske det komplekse forholdet mellom mediesentiment og markedsvolatilitet, og dekker ulike aspekter av denne dynamikken. Gjennom tre forskningsartikler presenterer avhandlingen en detaljert undersøkelse av hvordan medier – alt fra tradisjonelle nyheter til sosiale medieplattformer – former markedsadferd og sentimenter, og til slutt påvirker markedsvolatiliteten. I den første artikkelen benytter vi en ny tilnærming for å kvantifisere mediesentimentets innvirkning på markedsvolatilitet på tvers av ulike finansmarkeder. Ved å bruke avanserte teknikker for naturlig språkbehandling, trekker vi ut semantisk sentiment fra nyhetsoverskrifter og innlegg på sosiale medier. Våre funn avslører en tidsvarierende forbindelse mellom mediesentiment og markedsvolatilitet, som indikerer hvordan endringer i mediedrevet sentiment kan føre til svingninger i markedsvolatilitet. I den andre artikkelen skifter vi fokus til den prediktive kraften av sentiment i finansmarkedene. Vi undersøker rollen til mediesentiment, hentet fra både nyheter og sosiale medier, i å forutsi oljeprisvolatilitet, gitt dens kritiske rolle i den globale økonomien. Ved å bruke en hybridmodell som integrerer sentimentanalyse med prognosemodeller, avslører vi de distinkte prediktive evnene til nyhets- versus sosiale mediesentiment. Analysen viser at inkorporering av mediesentiment betydelig forbedrer nøyaktigheten av disse prediksjonene. I den tredje artikkelen utforsker vi informasjonstransmisjonsmønsteret fra ekte og falske nyheter til markedsvolatilitet. Gjennom en avansert analyse av politiske nyhetshistorier klassifiserer vi nyhetselementer som enten ekte eller falske og måler deretter deres respektive påvirkninger på aksjemarkedsvolatilitet. Vår studie avdekker de distinkte volatilitetsmønstrene assosiert med hver type nyhet, og fremhever den mer uttalte og samtidig kortvarige innflytelsen av falske nyheter sammenlignet med den vedvarende, men jevnere innflytelsen av ekte nyheter. | en_US |
dc.description.doctoraltype | ph.d. | en_US |
dc.description.popularabstract | This thesis explores the impact of news and social media on financial markets using advanced AI techniques. Paper I uncovers how fluctuations in media sentiment can lead to significant changes in market behavior, measuring the dynamic relationship between sentiment and market volatility with innovative methods. Paper II shows that by integrating sentiment analysis from both news and social media, traditional news is a more reliable forecaster of oil price volatility compared to social media information. Paper III reveals that while fake news triggers short-term market spikes, non-fake news has a more sustained and stable connection with market volatility. This study equips investors and policymakers with insights into how media sentiment shapes market movements. It emphasizes the crucial role of credible information and captures the influence of misinformation. By integrating AI with financial analysis, the study offers innovative tools for better understanding the role media information plays in the financial market. | en_US |
dc.identifier.isbn | 978-82-8266-267-3 | |
dc.identifier.uri | https://hdl.handle.net/10037/35326 | |
dc.language.iso | eng | en_US |
dc.publisher | UiT The Arctic University of Norway | en_US |
dc.publisher | UiT Norges arktiske universitet | en_US |
dc.relation.haspart | <p>Paper I: Abdollahi, H., Fjesme, S.L. & Sirnes, E. (2024). Measuring market volatility connectedness to media sentiment. <i>The North American Journal of Economics and Finance, 71</i>, 102091. Also available in Munin at <a href=https://hdl.handle.net/10037/34504>https://hdl.handle.net/10037/34504</a>.
<p>Paper II: Abdollahi, H. (2023). Oil price volatility and new evidence from news and Twitter. <i>Energy Economics, 122</i>, 106711. Also available in Munin at <a href=https://hdl.handle.net/10037/29978>https://hdl.handle.net/10037/29978</a>.
<p>Paper III: Abdollahi, H., Fjesme, S.L. & Sirnes, E. Fake news and market volatility: Insights from a large language model. (Manuscript). | en_US |
dc.rights.accessRights | embargoedAccess | en_US |
dc.rights.holder | Copyright 2024 The Author(s) | |
dc.rights.uri | https://creativecommons.org/licenses/by-nc-sa/4.0 | en_US |
dc.rights | Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) | en_US |
dc.subject | Market and media sentiment, Transmission mechanism, Machine learning in finance | en_US |
dc.title | Market volatility and new evidence from media sentiment: An AI-driven approach | en_US |
dc.type | Doctoral thesis | en_US |
dc.type | Doktorgradsavhandling | en_US |