PROJECT TITLE :

A Deep Neural Network for Crossing-City POI Recommendations

ABSTRACT:

Large amounts of location-based social media data, such as check-ins, are generated as a result of the proliferation of devices that are aware of their location (for example, smart phones). This motivates a great deal of research into point-of-interest recommendation methods, which utilize machine learning approaches. However, the majority of the studies that are currently available only make recommendations for POIs located within the same city or region. These studies do not make any recommendations for POIs that users can visit when they travel to a new city. In this paper, we propose a novel deep neural network for recommending Points of Interest (POI) when traveling between cities. We call our network ST-TransRec. The deep neural network, the transfer learning technique, and the density-based resampling method are all incorporated into a single, unified structure by this method. Deep neural networks are used in this model to learn the embeddings of points of interest (POIs) and to learn the preferences of users regarding POIs. In addition, the transfer learning method is utilized in order to close the knowledge gap that exists between cities as a direct result of the city-dependent characteristics. We design a density-based spatial resampling model because the distributions over POIs are uneven. This model makes it possible for POIs in different cities to be well matched with one another. Extensive tests are run on two different datasets taken from the real world. The results of the experiments demonstrate the benefits of using ST-TransRec rather than the methods that are considered to be state-of-the-art for making crossing-city POI recommendations.