Are sustainable cities attractive to digital nomads? A machine learning approach

This study explores the determinants of urban sustainability for digital nomads by predicting the nomad_score, a composite indicator of city attractiveness for remote workers, using supervised machine learning. Data were collected from NomadList and open city datasets, covering 668 destinations and 27 quantitative variables related to cost of living, connectivity, safety, social inclusion, and lifestyle. Gradient boosting, random forest, neural network, decision tree, AdaBoost, linear regression, and partial least squares (PLS) were evaluated using 20-fold cross-validation and metrics such as mean squared error (MSE), root MSE (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), and R². Our results show that ensemble methods based on boosting outperform other modes, with the best results being reported with AdaBoost (R² = 0.50 in training and 0.68 in prediction). The analysis of global predictors revealed that gender equality and entrepreneurship (startup_score and female_friendly) were the main features, whereas SHAP-based (SHapley Additive exPlanations) local explanations showed that Airbnb cost and safety were the most important ones. Policy implications point to the need to strengthen green infrastructure, expand digital connectivity, and promote social equity to make cities more attractive in the new remote-work economy.