Airbnb Price Prediction in Lisbon: A Machine Learning Approach to Price Estimation in Short-Term Rental Market

The rapid expansion of the short-term rental sector has significantly altered the urban and tourism landscape of the city of Lisbon, generating new economic opportunities alongside regulatory and housing sustainability challenges. In this respect, it is important to better understand the determinants for price formation and to develop interpretable predictive models to achieve better efficiency and fairness in this domain. This research will implement machine learning techniques to estimate the price of short-term rentals in the city of Lisbon, using public datasets from the Inside Airbnb site. The research follows the CRISP DM process model structured around data understanding, preparation, modelling and evaluation activities. The dataset merges thousands of listings with structural, geographic, reputational, and contextual variables. After preprocessing, transformation and generating derived variables which included distance to the city centre, metro stations, tourist attractions and premium areas, several forms of supervised learning algorithms were developed and compared across several versions of Random Forest, XGBoost and LightGBM. Parameters were optimised through a Bayesian approach to capture the trade-off between accuracy and generalisation, and model validation was achieved through k fold cross validation to ensure robustness of the performance assessment. The SHAP method ensured model interpretability by allowing for the understanding of the individual contribution of each variable and providing a transparency behind the behaviour of the model. Practically, this research is a foundation of empirical evidence to assist hosts to establish a more consistent strategy for pricing, digital platforms improve their recommendations and support public authorities to provide analytical evidence for recommendations about urban regulations. We showed, in this sense, that machine learning and interpretability have potential for enabling a deeper, more transparent, and sustainable comprehension of the short-term rental market in Lisbon.