An LSTM based Personalized Travel Route Recommendation System

With the rapid growth of the economy, more and more people are willing to spend their budget on travel. Due to the rapid development of social media, more people prefer self-guided tours to package itineraries planned by travel agencies. To fulfill tourists' personalized needs, a personalized travel route recommendation system based on social media is proposed. First, to provide meaningful route suggestions, we cluster nearby geotagged locations of photos into landmarks using the HDBSCAN (Hierarchical Density-Based Spatial Clustering of Applications with Noise) algorithm. Second, for each landmark, all hashtags marked in a landmark form a virtual document. The Latent Dirichlet Allocation (LDA) method extracts landmark topics from virtual documents. Third, the visiting time of each landmark and travel time between landmarks are obtained. Finally, a long-short term memory (LSTM) based model is proposed to predict a user's next visiting landmark and its visiting time. For users who do not have previous visiting records, our system can still generate an appropriate route according to the user's preference for topics. A set of case demonstrations and experiments show that the proposed system is sensitive to the users' input and performs better than the one without considering visiting time.