Handling Weak Causal Strength in Event Sequence Causal Discovery

Develop effective strategies for learning causal structure among event types from multi-type event sequences that can reliably detect causal relationships with weak causal strength, thereby addressing the low-recall issue observed under short observation periods when applying the Topological Hawkes Process-based framework.

Background

In the real-world metropolitan cellular network experiments, the authors observed that recall was limited, attributing this to the short observation period (one week) and the presence of causal relationships with relatively weak causal strength that are not detected by current methods.

They explicitly state that addressing weak causal strength remains an open problem, indicating the need for methodological advances that can improve sensitivity to low-strength causal effects within the Topological Hawkes Process framework or related causal discovery approaches on event sequences.