Artificial Intelligence for Trading Strategies
Details
- Authors: Roman Linder
- Title: Artificial Intelligence for Trading Strategies
- Supervisior: Prof. Dr. Jörg Osterrieder
- Degree: Bachelor of Science
- University: University of Applied Sciences Bern, BSc Business Information Systems
- Year: 2023
- Status: Working Paper
Summary
The bachelor thesis aimed to predict the S&P 500 stock market index using artificial intelligence (AI) and machine learning (ML). It also sought to analyze the impact of three additional factors (Treasury Yield 10 Year, CBOE Volatility Index, US Dollar/USDX Index) on improving the accuracy of the predictions. The study involved a literature review on AI in finance and a comprehensive investigation using a well-documented ML pipeline. Five Random Forest models were created and trained with different combinations of features. The comparison of these models revealed that integrating the additional factors significantly improved predictive accuracy by 12.5%. Volatility was found to be the most influential factor in predicting the S&P 500. The research highlights the effectiveness of utilizing AI and ML in stock price forecasting when careful attention is given to each step of the process. It suggests future research should continue to combine AI and ML methods with human expertise for the best results.
Abstract
The bachelor thesis "Artificial Intelligence for Trading Strategies: Forecast the S&P 500 and analyze the added value of additional factors in predicting this stock market index" set the goal of predicting the S&P 500 using artificial intelligence (AI) and machine learning (ML) and analyze the added value of three additional factors (Treasury Yield 10 Year, CBOE Volatility Index, US Dollar/USDX Index). Given the complexity of predicting stock prices, the work evaluated the value and effectiveness of integrating these additional elements in improving the predictive accuracy of the S&P 500 index.
The methodology included a literature review on Artificial Intelligence in the financial sector and a thorough investigation using an entirely constructed and documented ML pipeline. Five Random Forest models were created and trained with different combinations of features from the S&P 500 and the additional factors. The comparison of these models allowed for an informed discussion on the accuracy of the predictions and the added value of the additional factors.
The empirical findings of this research offer persuasive evidence that the inclusion of these additional factors is paramount in the accurate prediction of the S&P 500 Index. The substantial role of feature engineering is of particular interest, as it provides the model with the requisite information for making robust predictions. The study revealed that integrating the three additional factors engendered a 12.5% improvement in predictive accuracy. Furthermore, among all the examined factors, volatility emerged as the most influential in its bearing on the prediction of the S&P 500 Index.
This research underscores the efficacy of leveraging AI and ML methodologies in stock price forecasting, contingent upon meticulous attention to each phase of the process. This ensures the resultant models' accuracy and reliability. Future research endeavors are recommended to persist with the synergistic integration of AI and ML, preferably in conjunction with human expertise, to yield the most beneficial outcomes.
Important links
https://drive.google.com/drive/u/0/folders/1e_OY4755spQ95hvH4mIPkg8legpDDTrF