Development of Context-Aware Recommenders of Sequences of Touristic Activities

Identifier:  TDX:4239

Handle: https://hdl.handle.net/20.500.11797/TDX4239

Authors:  Orama, Ayebakuro Jonathan

Abstract:
In recent years, recommender systems have become ubiquitous on the web. Many web services, including movie streaming, web search and e-commerce, use recommender systems to aid human decision-making. Tourism is one industry that is highly represented on the web. There are several web services (e.g. TripAdvisor, Yelp) that benefit from integrating recommender systems to aid tourists in exploring tourism destinations. This has increased research focused on improving tourism recommender systems and solving the main issues they face. This thesis proposes new algorithms for tourism recommender systems that learn tourist preferences from their social media data to suggest a sequence of touristic activities that align with various contexts and include affine activities. To accomplish this, we propose methods for identifying tourists from their frequent Twitter posts, identifying the activities experienced in these posts, and profiling similar tourists based on their interests, contextual information, and activity periods. User profiles are then combined with an association rule mining algorithm for capturing implicit relationships between points of interest apparent in each profile. Finally, a rule ranking and activity selection process produces a set of recommendable activities. The recommendations were evaluated for accuracy and the effect of user profiling. We further order the set of activities using a multi-objective algorithm to enrich the tourist experience. We also carry out a second-stage analysis of tourist flows at destinations which is beneficial to destination management organisations seeking to understand tourist mobility. Overall, the methods and algorithms proposed in this thesis are shown to be useful in various aspects of tourism recommender systems.