MDP-based Itinerary Recommendation using Geo-Tagged Social Media

Planning vacations is a complex decision problem. Many variables like the place(s) to visit, how many days to stay, the duration at each location, and the overall travel budget need to be controlled and arranged by the user. Automatically recommending travel itineraries would thus be a remedy to quickly converge to an individual trip that is tailored to a user's interests. While on a trip, users frequently share their experiences on social media platforms e.g., by uploading photos of specific locations and times of day. Their uploaded data serves as an asset when it comes to gathering information on their journey. In this paper, we leverage social media, more explicitly photo uploads and their tags, to reverse engineer historic user itineraries. Our solution grounds on Markov decision processes that capture the sequential nature of itineraries. The tags attached to the photos provide the factors to generate possible configurations and prove crucial for contextualising the proposed approach. Empirically, we observe that the predicted itineraries are more accurate than standard path planning algorithms.