Abstract

In financial markets, stock price forecasting plays a critical role in investment decision-making, especially for globally influential companies like Apple Inc. This study aims to develop and assess models for predicting Apple Inc.'s stock price using various approaches, including regression analysis, smoothing methods, and the Box-Jenkins methodology. We analyzed ten years of Apple Inc.'s historical adjusted closing price data to construct models such as unregularized regression, regularized regression (Ridge and Lasso), smoothing methods (including exponential smoothing and moving averages), and the Box-Jenkins (SARIMA) model. The dataset was divided into training and test sets, and the predictive performance of each model was evaluated using the Average Prediction Squared Error (APSE). The findings indicate that the Simple Exponential Smoothing model performed best for short-term predictions, with an APSE of 0.01455. The Box-Jenkins model achieved an APSE of 0.08270, unregularized regression 0.021, while Ridge and Lasso models yielded APSEs of 0.6 and 4.97, respectively. In summary, smoothing methods are well-suited for short-term forecasting, while the Box-Jenkins method offers greater stability but comes with added complexity. Investors should choose forecasting models based on their specific requirements. This study provides empirical evidence for stock price forecasting and contributes new insights into the application of financial analysis and data science techniques.

Keywords: 

stock price forecasting, time series analysis, regression models, smoothing techniques, Box-Jenkins (SARIMA) Model

Issue

How to Cite