Comparative Analysis of Models for Apple Stock Price Prediction

Abstract

Predicting stock prices remains a difficult task because financial markets are influenced by many uncertain and rapidly changing factors. Traditional statistical models often fail to capture dynamic market patterns, while machine learning and deep learning approaches have demonstrated stronger predictive capabilities. This study investigates the performance of four forecasting models—ARIMA, XGBoost, Long Short-Term Memory (LSTM), and Transformer—for Apple Inc. (AAPL) stock price prediction using 20 years of historical daily data from Yahoo Finance. The closing price is used as the primary feature, and Root Mean Square Error (RMSE) serves as the evaluation metric. According to the experimental results obtained in this study, deep learning models provide lower prediction errors than traditional statistical approaches on the selected dataset. Among all models, the Transformer achieves the lowest RMSE and the best trend fitting. The findings demonstrate the effectiveness of AI techniques in financial time series forecasting and provide references for future research.