Predicting Domestic Tourists' Length of Stay in Italy leveraging Regression Decision Tree Algorithms

This study innovates in predicting domestic tourists' Length of Stay (LoS) in Italy by using decision tree models, addressing the gap in understanding LoS's determinants, and improving upon inconsistent results from traditional parametric analyses. Utilizing the 2019 "Viaggi e Vacanze" survey by the Italian National Institute of Statistics and categorizing variables into sociodemographic, economic, travel-related, and psychological factors, the research applies one-hot encoding to analyse 48,410,000 trips. Through evaluating random forest and gradient boosting models, the study highlights their superiority in identifying complex data patterns, offering actionable insights for tourism policymakers. These models enable precise LoS estimation, facilitating enhanced strategic planning for extending stays, optimizing services, and improving promotional efforts to maximize tourism's economic impact. This approach offers a comprehensive tool for developing policies that boost visitor engagement and economic benefits, showcasing a significant advancement in tourism management practices.‎