Year: 2025
Type: article
Abstract: The stock market is inherently volatile and influenced by a multitude of dynamic, interrelated factors ranging from macroeconomic indicators and geopolitical events to investor sentiment and technological innovation. For investors and financial analysts, accurately predicting stock prices remains a persistent challenge, particularly in the technology sector, where companies like Apple, Microsoft, Google, and Netflix experience rapid growth, frequent innovation cycles, and high market sensitivity. This project applies machine learning to predict future stock prices for five major tech firms, which include Apple, Tesla, Amazon, Google, and Microsoft. Past stock price datasets available on Kaggle were used, and by leveraging historical adjusted closing prices alongside engineered inputs such as daily returns, moving averages, rolling volatility, and lagged values. We train and compare sequential models, including TCNs, Informer architectures, and LSTMs. The analysis revealed several important insights about the current state of machine learning-based stock prediction. Model performance limitations were identified, with modest accuracies ranging from 49% to 63% for predicting whether stock prices would increase by more than 1.5% within five days. Stock-specific predictability and market insights were also highlighted, with Microsoft emerging as the most predictable (63% accuracy) due to its relatively stable price movements, while Tesla proved most challenging to forecast (49-50% accuracy). Our findings demonstrate that thoughtful feature construction combined with advanced sequence modeling can uncover market dynamics and boost prediction accuracy, showcasing the tangible benefits of AI-driven analytics for informed financial decision-making.
Authors: Tayo P. Ogundunmade, Olayinka B. Ayeni
Institution: University of Ibadan
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: gold
OpenAlex ID: W4415954329