Stock price prediction using machine learning and NLP techniques

Thumbnail Image




Καρπούζης, Σπυρίδων

Journal Title

Journal ISSN

Volume Title



This thesis investigates the application of Deep Learning and Natural Language Processing (NLP) techniques in predicting the stock prices of the Magnificent Seven stocks, namely Apple (AAPL), Microsoft (MSFT), Amazon (AMZN), Alphabet (GOOGL), Meta Platforms (META), Tesla (TSLA), and Nvidia (NVDA). The primary objectives are to demonstrate the superior performance of deep learning methods, particularly a novel architecture combining Generative Adversarial Networks (GAN) with NLP, over traditional approaches, to evaluate the impact of incorporating NLP data on prediction accuracy and to address the implications for the Efficient Market Hypothesis (EMH), which posits that stock prices fully reflect all available information. A quantitative methodology was employed, utilizing stock price data sourced from Yahoo Finance. The study finds that models based on Long Short-Term Memory (LSTM) networks and GANs, when integrated with NLP data from tweets, outperform traditional methods such as the AutoRegressive Integrated Moving Average (ARIMA). However, the performance of these models varies according to the specific stock and its volatility, suggesting that different models may be optimal under different conditions. These results highlight the potential benefits of advanced deep learning models, particularly those augmented with NLP, in enhancing the accuracy of stock price prediction. Moreover, the findings contradict the EMH by demonstrating that our models achieved statistically significant improvements in prediction accuracy. The findings contribute to the growing body of knowledge on the application of AI and Machine Learning in financial markets and suggest directions for future research.



Stock prediction, Deep learning, Long short-term memory, Generative adversarial networks, Natural language processing