Predicción de la demanda estacional de productos en una Comercializadora de juguetes mediante modelos de series eemporales.

Abstract

Abstract: This study aims to forecast the seasonal demand of a toy retail company located in the Historic Center of Quito by using a SARIMA econometric time series model, with the purpose of improving demand anticipation and supporting commercial planning decisions. Weekly historical sales data covering a 72-month period from January 2020 to October 2025 were analyzed. The CRISP-DM methodology was applied, including business understanding and an exploratory analysis of the time series. A Box-Jenkins modeling approach was implemented, incorporating the Augmented Dickey-Fuller test to assess stationarity, as well as autocorrelation and partial autocorrelation functions to identify the temporal structure of the process. Based on these results, a SARIMA (1,0,3)(1,1,1,52) model was estimated, successfully capturing peak demand periods and annual seasonality associated with the commercial cycles of the sector. Diagnostic tests indicated that residuals behaved as white noise and followed an acceptable normal distribution, while AIC and BIC criteria showed an appropriate balance between model fit and complexity. Overall, the results confirm the validity of the model and its capacity to reproduce the essential structure of the series, enabling useful forecasts to anticipate high-demand periods and reduce uncertainty in inventory planning.