How can Natural Language Processing techniques help predict the NASDAQ Stock Market? / Aurélie GUERIN / 2021

Titre : How can Natural Language Processing techniques help predict the NASDAQ Stock Market? Type de document : Mémoire Auteurs : Aurélie GUERIN, Auteur Année de publication : 2021 Importance : 58 p. Langues : Anglais (eng) Mots-clés : Entreprise BNP PARIBAS Management FINANCEMENT Résumé : Background – The recent advancement of Artificial Intelligence, particularly in the finance industry, has led to the emergence of Natural Language-based Financial Forecasting. However, studies on NLP based predictions of the NASDAQ Stock Market have not received adequate attention. Objective – The objective of this study was therefore to investigate NLP techniques enabling financial predictions of the NASDAQ Stock Market using various sources of textual data combined with different approaches for Sentiment Analysis. Method – To conduct our study, we collected open-sources datasets composed of Reuters financial news on one side, and tweets on the other side. We also collected time series data of the NASDAQ stock market. We then preprocessed the textual data. This step was absolutely crucial to ensure the format could be handled by the algorithm. Afterwards, we proceeded with determining the polarity score of the texts in order to perform Sentiment Analysis. Results – We succeeded in creating a model that yields both higher returns than a passive buy and hold strategy while also providing lower volatility. It used the body of Reuters financial news as data source and employed a dictionary-based approach for Sentiment Analysis. Conclusion – This study will contribute to the investors’ knowledge on how to achieve a performing strategy using NLP techniques. Nevertheless, further research concerning the use of news coming from other sources, combination of more tweets’ characteristics and in other languages are needed. Note de contenu : Pour accéder aux fichiers PDF, merci de vous identifier sur le catalogue avec votre compte Office 365 via le bouton CONNEXION en haut de page. Programme : MS Marketing & Data Analytics Permalink : https://cataloguelibrary.neoma-bs.fr/index.php?lvl=notice_display&id=539110