Tourism demand forecasting in Portugal's municipalities

: an explainable machine learning approach

Abstract

In recent decades, the tourism industry has experienced remarkable growth, making it essential to provide accurate forecasts of tourism demand for the efficient allocation of resources and continued growth of the industry. This thesis presents a machine learning approach to forecast tourism demand in Portugal. However, forecasting in the tourism sector faces a challenge: compromising the balance between model accuracy and interpretability. While some highly accurate models lack transparency, making them difficult to understand, this study addresses this concern using the Tree SHAP method. By identifying the contributions of the features, this approach offers a globally interpretable model, increasing the reliability of tourism demand forecasts. This thesis aims to answer: 1) What are the tourism demand trends in Portugal?; 2) What are the key features that most significantly influence tourism demand for different tourist accommodations in Portugal's municipalities?; 3) How to trigger valuable insights in tourism demand forecasting models via the application of explainability strategies? For this purpose, non-public data from Turismo de Portugal and additional variables, such as population, are used to train an XGBoost model. The main predictors of demand in Portugal were revealed, including summer, population, and number of beds. This study has practical implications for policymakers and management teams in the marketing and tourism sectors, providing valuable information for decision-making in the sector.

Date of Award	30 Jan 2024
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Ana Guedes (Supervisor)