The repository contains code for end-to-end training and testing of a transformer based Hierarchical Graph Neural Network that performs next-POI prediction on time series extracted from the New York foursquare dataset (dataset 2 here).
Our implementation beats the one presented in the paper on all reported metrics, a detailed description of the methods used and the results achieved can be found in the attached report here.