ABSTRACT
Most hotel recommendation systems currently rely on text-based information or meta-data. We develop a deep network recommendation model with three modalities – picture, review, and scoring .We propose a unifified deep neural network including an embedding layer, pooling layer, and fully connected layer. Comparing with other algorithms, we verify its efficacy in improving travel recommendations based on the hotel data crawled from Ctrip and the major evaluation indicators. Our study contributes to the literature by building a knowledge model for tourist hotels based on the analysis of user-generated data and providing practical guidance for hotel managers and users.
Disclosure statement
No potential conflict of interest was reported by the author(s).