Attribute-Based Conformance Diagnosis: Correlating Trace Attributes with Process Conformance

An important practical capability of conformance checking is that organizations can use it to alleviate potential deviations from the intended process behavior. However, existing techniques only identify these deviations, but do not provide insights on potential explanations, which could help to improve the process. In this paper, we present attribute-based conformance diagnosis (ABCD), a novel approach for correlating process conformance with trace attributes. ABCD builds on existing conformance checking techniques and uses machine learning techniques to find trace attribute values that potentially impact the process conformance. It creates a regression tree to identify those attribute combinations that correlate with higher or lower trace fitness. We evaluate the explanatory power, computational efficiency, and generated insights of ABCD based on publicly available event logs. The evaluation shows that ABCD can find correlations of trace attribute combinations with higher or lower fitness in a sufficiently efficient way, although computation time increases for larger log sizes.