Bitte benutzen Sie diese Kennung, um auf die Ressource zu verweisen:
https://doi.org/10.21256/zhaw-24410| Publikationstyp: | Thesis: Master |
| Titel: | Natural Language Processing in finance : analysis of sentiment and complexity of news and earnings reports of swiss SMEs and their relevance for stock returns |
| Autor/-in: | Schlaubitz, Alexander |
| Betreuer/-in / Gutachter/-in: | Fazlija, Bledar |
| DOI: | 10.21256/zhaw-24410 |
| Umfang: | 94 |
| Erscheinungsdatum: | 2021 |
| Verlag / Hrsg. Institution: | ZHAW Zürcher Hochschule für Angewandte Wissenschaften |
| Verlag / Hrsg. Institution: | Winterthur |
| Sprache: | Englisch |
| Fachgebiet (DDC): | 006: Spezielle Computerverfahren 332: Finanzwirtschaft |
| Zusammenfassung: | Natural Language Processing (NLP) can be used for the purpose of achieving human-like lan-guage interpretation for a range of tasks such as analyzing stock related news and earnings re-ports. This allows financial professionals to gain a quicker and deeper insight into market infor-mation via Machine Learning (ML) algorithms. In literature, significant connections between textual sentiment and stock prices for large cap companies in the English-speaking world are identified. This paper examines if such effects also pertain to Swiss companies that create their reporting in the German language. Accordingly, the interplay of sentiment and complexity of news and earnings reports for publicly listed Swiss companies in the small and mid cap segment are analyzed, together with their potential relevance for stock returns. Additionally, the Trans-former algorithm is introduced, and its performance is compared to older ML algorithms. A literature review showed that the Transformer model was able to set new state-of-the-art per-formance in language modeling and thus outperform previous language-oriented algorithms. Accordingly, for the sentiment analysis of Swiss earnings report and news articles, a DistilBERT model was trained and fine-tuned using a financial phrase databank. The model, which made use of Transfer Learning, showed sentiment prediction accuracies of 90 percent. Text complexity was analyzed using the well-established Flesch Score and Wiener Sachtextformel. The results for complexity show that both, Swiss earnings reports and financial news articles, are generally dif-ficult to understand and require the reading level of university students. Sentences with either negative or positive sentiment are both equally complex and there is not any noticeable differ-ence in complexity between news and earnings reports, at least for human readers. For sentiment it was found that only a small number of earnings reports contains negative sentiment on an ag-gregated level, even when firms reported lower earnings. Meanwhile, news articles provide a more balanced data source for sentiment. Significant linear connections in terms of regression analysis of the predictive ability of text sentiment on future stock returns were found for only 4 of total 15 analyzed companies, indicating that sentiment alone is a bad linear predictor for future performance. However, a nonlinear classification model found an increase of 8 percent in accu-racy when including sentiment data together with historical stock data, compared to only using historical stock data for the forecast of stock return development. |
| URI: | https://digitalcollection.zhaw.ch/handle/11475/24410 |
| Lizenz (gemäss Verlagsvertrag): | CC BY-NC-ND 4.0: Namensnennung - Nicht kommerziell - Keine Bearbeitungen 4.0 International |
| Departement: | School of Management and Law |
| Enthalten in den Sammlungen: | MSc Banking and Finance |
Dateien zu dieser Ressource:
| Datei | Beschreibung | Größe | Format | |
|---|---|---|---|---|
| Masterarbeit_Alexander_Schlaubitz.pdf | 1.87 MB | Adobe PDF |  Öffnen/Anzeigen |
Zur Langanzeige
Schlaubitz, A. (2021). Natural Language Processing in finance : analysis of sentiment and complexity of news and earnings reports of swiss SMEs and their relevance for stock returns [Master’s thesis, ZHAW Zürcher Hochschule für Angewandte Wissenschaften]. https://doi.org/10.21256/zhaw-24410
Schlaubitz, A. (2021) Natural Language Processing in finance : analysis of sentiment and complexity of news and earnings reports of swiss SMEs and their relevance for stock returns. Master’s thesis. ZHAW Zürcher Hochschule für Angewandte Wissenschaften. Available at: https://doi.org/10.21256/zhaw-24410.
A. Schlaubitz, “Natural Language Processing in finance : analysis of sentiment and complexity of news and earnings reports of swiss SMEs and their relevance for stock returns,” Master’s thesis, ZHAW Zürcher Hochschule für Angewandte Wissenschaften, Winterthur, 2021. doi: 10.21256/zhaw-24410.
SCHLAUBITZ, Alexander, 2021. Natural Language Processing in finance : analysis of sentiment and complexity of news and earnings reports of swiss SMEs and their relevance for stock returns. Master’s thesis. Winterthur: ZHAW Zürcher Hochschule für Angewandte Wissenschaften
Schlaubitz, Alexander. 2021. “Natural Language Processing in Finance : Analysis of Sentiment and Complexity of News and Earnings Reports of Swiss SMEs and Their Relevance for Stock Returns.” Master’s thesis, Winterthur: ZHAW Zürcher Hochschule für Angewandte Wissenschaften. https://doi.org/10.21256/zhaw-24410.
Schlaubitz, Alexander. Natural Language Processing in Finance : Analysis of Sentiment and Complexity of News and Earnings Reports of Swiss SMEs and Their Relevance for Stock Returns. ZHAW Zürcher Hochschule für Angewandte Wissenschaften, 2021, https://doi.org/10.21256/zhaw-24410.
Alle Ressourcen in diesem Repository sind urheberrechtlich geschützt, soweit nicht anderweitig angezeigt.