NPR-LBN: next point of interest recommendation using large bipartite networks with edge and cloud computing

Abstract During the last decades, tourism has been augmented worldwide through which the diversity of tourists' interests is increased and is challenging to tackle with the traditional management system. Such challenges can be overcome by LBSNs (Location-Based Social Networks) such as Yelp, Foursquare, and Facebook which help to collect more personalized information close to tourists' preferences/interests like check-ins, comments, and reviews. In this regard, solutions have been proposed to exploit the POI (Point of Interest) recommendation, but they failed to overcome sparsity and cold-start problems. Existing methods are also not focusing on important aspects, including geographical context, dynamics preferences and social influence, which are essential factors in POI recommendation. Therefore, this work tried to incorporate these factors and present a unified model using bipartite networks to learn users and POI dynamics. For this purpose, we have represented all the factors using eleven networks and combined them into a single latent space. In addition, Edge Computing processes data at the network's edge, reducing latency and bandwidth usage and enabling real-time and personalized recommendations. Furthermore, cloud computing could be used to store and process the large amounts of data collected from LBSNs, to support the proposed model's computational requirements and make it more accessible and scalable, allowing it to be easily used by tourism management systems worldwide. Experimental results show that our model outperforms state-of-the-art methods using real-world dataset in terms of accuracy and perform better against sparsity and cold-start problems.