Personalized travel route recommendation from multi-source social media data

• DOI: 10.1007/s11042-018-6776-9
• Publication Date: 2018-10-18T09:29:24Z
• Authors: Gang Hu; Yi Qin; Jie Shao
• Keywords: Sequence (biology); Fuse (electrical); Similarity (geometry); Point of interest

Personalized travel route recommendation aims to recommend tourist attractions based on user’s interest and generate routes which can be viewed as a recommendation and sequence task. How to build interest model for users to find personalized attractions and generate the location sequence brings great challenges. In this work, multi-source social media (e.g., travelogues and check-in records) are leveraged to find user interests and model route attributes. In order to fuse multi-source data in a unify metric space, a topical package is built as the measurement space. Then, a long short-term memory (LSTM) based method is used to generate some candidate positions based on sparse user-specified inputs. Finally, top ranked routes are recommended as final results. The proposed approach combining multi-source social based topical package and LSTM is evaluated on a real travel dataset and compared with three state-of-the-art methods. The experimental result shows that our method performs better for providing personalized travel routes.