Abstract

This study compares six approaches to predicting hotel occupancy based on six dimensions of perceived quality (value for money, location, comfort, room quality, cleanliness, and service) in a sample of 337 hotels. In general terms, machine learning (ML) algorithms outperform traditional models in predictive accuracy: Random Forest obtains R² ≈ 0.96, the decision tree R² ≈ 0.89, and R² ≈ 0.86. In contrast, OLS achieves an R² = 0.839, while the two fuzzy regression variants (Tanaka, and Diamond and Kloeden) register an R² of approximately 0.81, with central coefficients very close to those of OLS, but with the advantage of offering fuzzy intervals that explicitly quantify uncertainty. In other words, the superiority of ML algorithms in R² coexists with the explanatory and risk management strength of fuzzy regression, especially when the data come from subjective judgments and Likert scales. Therefore, a portfolio of models is recommended: AM for operational forecasting and fuzzy approaches for planning under uncertainty.

You must be a member to download the full article.