Comparative Analysis of Linear Regression and ARIMA Models for Stock Price Forecasting

Abstract

Stock price forecasting has long been a critical area in finance, as accurate predictions provide valuable guidance for investors, portfolio managers, and policymakers in navigating uncertain markets. Traditional statistical models remain widely used because of their interpretability and relatively low computational cost, even as more complex machine learning methods gain popularity. This study investigates the effectiveness of linear regression and Autoregressive Integrated Moving Average (ARIMA) models in forecasting stock prices. Using historical data from APPL, GOOG, TSLA, NVDA and MSFT, the models were evaluated under different market conditions. Results show that linear regression provides strong in-sample fitting, especially for stable stocks with lower volatility. However, it struggles with highly volatile stocks due to sudden market shocks. In contrast, ARIMA captures time-series dynamics more effectively and delivers more accurate short-term forecasts, especially in volatile environments. The findings suggest that both models have complementary strengths, and hybrid approaches may further improve stock price prediction in complex financial markets.