Seasonal ARIMA Model for Forecasting Microsoft's Revenues

Abstract

This study aims to analyze the quarterly revenue series of Microsoft Corporation using time series econometric techniques, with a focus on identifying patterns of seasonality and forecasting future values. The series was initially found to be non-stationary, and stationarity was achieved after first differencing. Seasonal components were confirmed through visual inspection of ACF and PACF plots, as well as through statistical criteria. Several SARIMA models were estimated and compared based on AIC, BIC, and the significance of coefficients. Among the evaluated models, SARIMA(2,1,2)(1,1,1)[4] was selected as the optimal model, exhibiting the lowest AIC and BIC values, with all parameters statistically significant. The residuals of the model were tested using the Ljung–Box Q-statistics and were found to be uncorrelated, indicating that the model sufficiently captured the dynamics of the data. The model was used to generate forecasts for future quarters through 2026, offering valuable insights for strategic planning and financial forecasting. The findings confirm the presence of both trend and seasonal patterns in Microsoft’s revenue series and demonstrate the effective