Deep Learning-based Mobile Tourism Recommender System

Abstract

A tourism recommendation system is a crucial solution to help tourists discover more diverse tourism destinations. A content-based approach in a recommender system can be an effective way of recommending items because it looks at the user's preference histories. For a cold-start problem in the tourism domain, where rating data or past access may not be found, we can treat the user's past-travel-photos as the histories data. Besides, the use of photos as an input makes the user experience seamless and more effortless. The current development in Artificial Intelligence-based services enable the possibilities to implement such experience. This research developed a Deep Learning-based mobile tourism recommender system that gives recommendations on local tourism destinations based on the user's favorite traveling photos. To provide a recommendation, we use cosine similarity to measure the similarity score between one's pictures and tourism destination's galleries through their label tag vectors. The label tag is inferred using an image classifier model that runs from a mobile user device through Tensorflow Lite. There are 40 label tags, which refer to local tourism destination categories, activities, and objects. The model is trained using state-of-the-art mobile deep learning architecture EfficientNet-Lite. We did several experiments and got an accuracy result of more than 85% on average, using EfficientNet-Lite as the base architecture. The implementation of the system as an Android application has been proved to give an excellent recommendation with Mean Absolute Percentage Error (MAPE) equals to 5%.