Akademik Veri Yönetim Sistemi
19th Conference on Computer Science and Intelligence Systems, FedCSIS 2024, Belgrade, Sırbistan, 8 - 11 Eylül 2024, ss.207-212
Airbnb, a prominent online marketplace, facilitates short- and long-term rentals by connecting customers with property owners offering entire apartments or private rooms. Accurate price prediction is crucial for both the platform and rental property owners. Previous studies have primarily focused on statistical methods and pre-processing techniques, with limited exploration of the impact of location attributes. This paper aims to enhance price prediction models for Airbnb listings by incorporating location data. Utilizing data from InsideAirbnb for Istanbul, we implemented various data pre-processing techniques and enriched the dataset with location-specific information. Our findings show that incorporating these location-based features significantly improved model performance, increasing the adjusted R2 metric by 22.5% and reducing Mean Absolute Error (MAE) by %10. This enhancement was achieved by using location-related index values and public transportation data provided by the Istanbul Metropolitan Municipality. These advancements can help property owners optimize rental prices and assist urban planners in making informed decisions about city infrastructure development.