On accurate POI recommendation via transfer learning

Point of interest (POI) recommendation is of great value for both service providers and users. However, it is hard due to data scarcity. To this end, in this paper, we propose a transfer learning based deep neural model, which fuses valueable cross-domain knowledge to achieve more accurate POI recommendation. We first learn the user’s spatial and non-spatial preferences based on their historical POI interactions. The model further captures user interactions in other domains and introduces useful preferences into POI recommendations, which can address data sparsity problems. Compared to the matrix factorization based cross-domain techniques, our method utilizes deep transfer learning, which can learn complex user-item interaction relationships and accurately capture user general preferences to transfer. Finally, we evaluate the proposed model using three real-world datasets. The experimental results show that our model significantly outperforms the state-of-the-art approaches for POI recommendation.