VoyageWithUs : a recommender platform that enhances group travel planning
MetadataShow full item record
Group travel planning poses unique challenges such as choosing hotels, restaurants and venues while catering to everyone’s wants and needs, or sharing trip itineraries and artifacts among trip participants. State of the art travel planning applications such as Yelp and TripAdvisor, while integrating with social networks and making recommendations, don’t offer recommendations for specific groups of travelers. On the other hand, while TripCase offers trip planning capabilities and email sharing, it doesn’t offer a full interactive travel planner that allows groups to contribute to the travel planning process. This report proposes an approach to making personalized group travel recommendations based on hybrid recommendation techniques that aggregates individual recommendations to find common ground between trip participants. This is achieved by designing a recommender system that uses data from a location based social network(LBSN) and makes recommendations based on the trip location, then refines them by applying incremental filters which are responsible for incorporating user preferences, similarity to other users and user context. Finally, it takes the generated recommendations for each trip participant and ranks them such that the items most highly ranked are the ones most likely to fit everyone’s preferences. The rationale for choosing a hybrid recommender system is to address common issues such as the cold start problem, where the quality of the recommendations is affected by either too few reviewers for a certain point of interest(POI) or too few reviews generated by trip participants. These issues, along with a coverage of related work is detailed in the first part of this report. In order to make the applicability of the recommender more tangible, I integrated it into a proof of concept mobile application that also allows travelers to collaborate and share travel planning artifacts, and generates itineraries based on the recommendations made. The recommender accuracy was measured against recommendations made by state of the art applications, while individual filters were evaluated using commonly used metrics. The recommender was tested in a series of relevant scenarios proving the effectiveness of the approach in making group travel recommendations, versus individual recommendations generated by other applications.