Forecasting The Closing Prices of Google and Microsoft with ARIMA Model

Abstract

Accurately forecasting stock prices is crucial for investors and policymakers since short-term errors can compound into costly decisions in volatile markets. This study uses an autoregressive integrated moving average (ARIMA) model and employed a 70/30 time series to test segmentation and rolling prediction design to analyze the daily closing prices of two leading technology companies in the United States, Alphabet Inc. (Google) and Microsoft Corporation. It implements both one-step-ahead and five-step-ahead rolling predictions to assess short- versus medium-horizon performance. The one-step setting closely tracks realized prices for both firms, yielding low absolute errors and particularly strong percentage accuracy for Microsoft; extending to five steps increases errors as expected, with Alphabet retaining comparatively better trend-following performance while Microsoft exhibits larger variance and slippage. These results suggest ARIMA remains a practical baseline for short-horizon operational decisions (e.g., daily risk controls, inventorying hedges) and for medium-term orientation where trends are smoother; however, horizon choice should reflect each asset’s volatility structure. The framework provides a transparent benchmark for subsequent hybrid or exogenous-variable models and can guide practitioners on when classical time-series tools suffice versus when richer models are warranted.