Uso de dados do Airbnb para predição do preço de aluguel de imóveis na cidade do Rio de Janeiro

Abstract

This study aims to apply machine learning algorithms to predict rental prices of properties listed on the Airbnb platform, focusing on the city of Rio de Janeiro. A public dataset provided by the platform itself was used, and preprocessing, exploratory data analysis, and variable selection were conducted to identify the features most correlated with price. Three regression models were trained and evaluated: Random Forest, XGBoost, and LightGBM. The Random Forest algorithm achieved the best performance, with a coefficient of determination (R²) exceeding 90%. The most relevant variables identified include accommodation type, number of bedrooms, bathrooms, beds, location, and offered amenities. The results demonstrate the potential of data science and machine learning techniques as effective tools for automated and accurate pricing of short-term rental properties. As future work, it is suggested to incorporate seasonal, geographic, and local event factors, and to develop an application that assists property owners in setting prices based on real-time data.