| dc.description.abstract | This thesis explores the influence of image features on the predictive performance of hedonic price models for Airbnb listings. By integrating machine learning methods, image quality features, colour features, and black-box model interpretation methods, the study demonstrates the value of these components in the field of property price prediction. This thesis utilizes a novel dataset scraped in 2023 from Amsterdam which offers updated insights into the role of image features in Airbnb pricing. After deploying 10 different machine learning models, the XGBoost model yields the best predictive accuracy based on several performance metrics. Although the enhancement in predictive performance of the XGBoost model by inclusion of image features was not statistically significant, these features showed non-negligible influences and interactions in the decision-making process of the model. These findings imply a potential role of image features in refining property price models, providing valuable insights for the stakeholders in the fields of hospitality, real estate, advertising, and machine learning research. | en_US |
