Host’s Assistant: Leveraging Graph Neural Networks for Daily Room Rate Prediction on Online Accommodation Sites

Pricing has always been an important topic. It takes a lot of effort and time from the most traditional economic point of view. Therefore, many researchers are devoted to the research of automatic pricing models. In our case, we are focused on the prediction of accommodation prices. Unlike real estate price prediction, we primarily study the pricing issues faced by Airbnb hosts on a daily basis. Currently, existing methods predict prices based on room factors. However, we believe that the relationship between neighboring rooms has a significant impact on prices, which has not been adequately considered. Additionally, there is a lack of flexible mechanisms that adjust prices based on demand forecasting and variations during peak and off-peak seasons. To address these challenges, we have developed an advanced Room Price Interactive Forecasting System (RPics). In addition to incorporating data on individual room characteristics, our primary focus is on the impact of neighboring rooms. We utilize the framework of the model to simulate the effects of mutual influence and forecast demand patterns for different days, including simulations for weekends or peak and off-peak travel seasons. The overall framework enables us to adjust prices accordingly and achieve superior results.