Implementing Machine Learning for Smart Tourism Frameworks

Considering that Artificial Intelligence is a game-changer in the smart tourism business, one of our key contributions is the formal presentation of frameworks that leverage AI technologies in the context of smart tourism. The utilisation of user-captured photographs in the smart tourism context is one novel approach shared by both frameworks that we believe will usher in a new era of smart tourism recommendations. We conceived the innovative concept “Moments of Interest” (MOIs), which is applied to a mobile application designed to provide personalised recommendations to tourists based on “moments” captured in user-taken photographs; as a result of this new and widely established behaviour of capturing images via smartphones. An important contribution of this book is an application that harvests and processes images captured by users in real-time and in the past to create a “Memories Database,” employing image labelling via machine learning and distributing the analysed data back to users via an application’s map-infused interface. The proposed revolutionary cloud-based crowdsourcing application for increasing smart tourism proposals utilises user-captured photos and context awareness to produce an innovative smart tourist experience. By isolating the image labelling module from the above-mentioned smart tourism application, we were able to analyse user-captured images that reside on tourists’ smartphones, collect and store the most frequent labels of touristic interest that reside in them on the cloud in both users’ profiles and a labelling table and then analyse these images. At the same time, photographs relating to POIs were labelled with a Deep Neural Network model in order to collect labels of tourist attractions pertinent to our POIs database. Furthermore, pre-trained Neural Network Matrix Factorization models were used to provide POI recommendations based on two distinct matrices: a user-POI rating matrix and a user-labels interaction matrix. In addition, the labelling table was used to locate a similar user to the target user if s/he has done a limited number of ratings or none at all. Thus, another key contribution of this book, provided in this section, is a framework that uses Deep Neural Networks to analyse several types of data, namely photographs and user-item interaction matrices, in order to realise smart tourist personalisation. This framework can be either a standalone application or a key component of future smart tourism applications because it requires minimal user data and interaction and leverages cutting-edge technologies for overcoming the cold start problem and the data sparsity problem while generating personalised recommendations from two distinct data sources.