Hyperbolic-Based Tree Edit Distance as Similarity of User Behavior

In this paper, we evaluate the similarity between users' behavior by applying tree edit distance (TED) to tree representations generated through hyperbolic metric learning methods. Understanding user behavior similarity is an essential task in e-commerce, as it can be modeled as a tree-like structure where each action corresponds to a node, representing the user's possible next move. Analyzing user behavior similarity through the lens of advertising costs provides insight into the effort needed to shift from one user's action pattern to another. We explore the benefits of using tree edit distance for user behavior analysis and further consider their similarity under hyperbolic metric embedding methods.