A link prediction-based recommendation system using transactional data

In this study, the research task is formulated as a link prediction problem. In this context, a heterogeneous graph that contains different types of nodes and links is generated from the transactions where nodes correspond to users and items, and links represent interaction occurrences between these users and items. Given such a network generated from past transactions, the goal is to discover new possible user-item link formations in a future network version. Since the problem investigates link occurrences between pairs of nodes, it can also be formulated as a binary classification task…