Predicting Prices of Tech Stocks using the Transformer
(2024) NEKH02 20241Department of Economics
- Abstract
- Accurately predicting future stock prices is a dream for many investors. The inherent uncertainty and complexity of financial markets makes price prediction a difficult task. However, the recent advancements of neural networks in the field of machine learning have introduced new possibilities for accurately predicting future prices. We explore the application of the Transformer, a recent model which has proven to have powerful performance on a number of tasks. By leveraging the Transformer, we seek to challenge the efficient market hypothesis. The hypothesis states that asset prices fully reflect all available information, making it impossible to consistently outperform the market by accurately predicting future prices. We demonstrate the... (More)
- Accurately predicting future stock prices is a dream for many investors. The inherent uncertainty and complexity of financial markets makes price prediction a difficult task. However, the recent advancements of neural networks in the field of machine learning have introduced new possibilities for accurately predicting future prices. We explore the application of the Transformer, a recent model which has proven to have powerful performance on a number of tasks. By leveraging the Transformer, we seek to challenge the efficient market hypothesis. The hypothesis states that asset prices fully reflect all available information, making it impossible to consistently outperform the market by accurately predicting future prices. We demonstrate the Transformers ability to relatively accurately predict future prices for five stocks, with a mean absolute percentage error of 3.18% on average. We contribute to an ongoing discussion on the validity of the efficient market hypothesis and how the Transformer can be leveraged for this task. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9158003
- author
- Zitouni, Théodore LU and Grull, Sebastian LU
- supervisor
- organization
- course
- NEKH02 20241
- year
- 2024
- type
- M2 - Bachelor Degree
- subject
- keywords
- Price Prediction, Neural Networks, Efficient Market Hypothesis, Machine Learning, Transformer
- language
- English
- id
- 9158003
- date added to LUP
- 2024-09-24 09:00:38
- date last changed
- 2024-09-24 09:00:38
@misc{9158003, abstract = {{Accurately predicting future stock prices is a dream for many investors. The inherent uncertainty and complexity of financial markets makes price prediction a difficult task. However, the recent advancements of neural networks in the field of machine learning have introduced new possibilities for accurately predicting future prices. We explore the application of the Transformer, a recent model which has proven to have powerful performance on a number of tasks. By leveraging the Transformer, we seek to challenge the efficient market hypothesis. The hypothesis states that asset prices fully reflect all available information, making it impossible to consistently outperform the market by accurately predicting future prices. We demonstrate the Transformers ability to relatively accurately predict future prices for five stocks, with a mean absolute percentage error of 3.18% on average. We contribute to an ongoing discussion on the validity of the efficient market hypothesis and how the Transformer can be leveraged for this task.}}, author = {{Zitouni, Théodore and Grull, Sebastian}}, language = {{eng}}, note = {{Student Paper}}, title = {{Predicting Prices of Tech Stocks using the Transformer}}, year = {{2024}}, }