Deep Learning Based Dynamic Pricing Model for Hotel Revenue Management

Abstract

Dynamic pricing, which determines optimal prices of products or services, is a research topic that has received a considerable amount of attention from different scientific communities: operations research and management science, marketing, economics, econometrics, and computer science. In this paper, we describe a dynamic pricing system for hotel revenue management. The proposed dynamic pricing system consists of three parts. The first one analyzes a hotel and its competitors so that we can set a reasonable base price. The second part forecasts future occupancy of a hotel, in this part we propose a novel sequence learning model for occupancy prediction, which integrates Deep Factorization-Machine (DeepFM) and the seq2seq model. Compared with several classical methods for time series prediction, our proposed model can significantly improve the performance of hotel occupancy forecasting without very complex feature engineering. The third part of the dynamic pricing system adopts a Deep Neural Network (DNN) to model human expertise and make adjustment for the hotel room price according to the base price, the predicted occupancy and other relating factors. We compared our dynamic pricing system with a rule based pricing strategy made by revenue management experts for evaluation, the results show that our model can suggest more rational price adjustment and make hotel revenue management more efficient.