Tourists’ presence flash estimation by Mobile Network Operators data

Abstract

The use of big data to anticipate the release of official statistics is attracting growing attention and interest within the scientific community. This paper presents a methodological contribution to flash estimation of overnight stays in Italian municipalities using administrative and mobile network operator (MNO) data. Three forecasting strategies: Simplistic, Quasi-Transfer Learning (QTL), and Augmented Learning (AL) are compared across different models (Linear, Random Forest, and Ensemble). The application focuses on Emilia-Romagna municipalities, evaluating predictive accuracy through repeated cross-validation. Results show that both QTL and AL outperform traditional methods, especially in terms of mean absolute percentage error (MAPE) and confidence interval stability.