Predicting Prices of Airbnb Accommodations in Thailand by SVM and XGBoost Methods

Abstract

In this study, our objective was to predict accommodation prices in Bangkok utilizing Airbnb data. The data went through necessary preparation procedures and was split into training and test sets. Both support vector machines and extreme gradient boosting methodologies were employed and optimized through hyperparameter tuning. However, the detection of overfitting necessitated a reassessment of feature selection. Several features were identified as having high importance values in both models, including the number of bedrooms, proximity to tourist destinations and landmarks in Bangkok, maximum property capacity, and the number of host listings. Additionally, support vector machines with the top 10 features outperformed other models, exhibiting the lowest mean absolute error (385.37) and root mean squared error (526.16) values. Crucially, features such as the number of bedrooms, proximity to tourist destinations, maximum property capacity, private room type, and provision of safety and facility information played significant roles. In conclusion, this study emphasizes the significance of machine learning in comprehending accommodation prices. The results highlight the importance of considering specific features, such as those identified, when setting accommodation prices.