Making Sense of Temporal Event Data:A Framework for Comparing Techniques for the Discovery of Discriminative Temporal Patterns

Extracting knowledge from complex data in an explicit formalization is one of the main challenges in creating human understandable descriptions of data, and in bringing humans in the loop when analyzing it. Recent developments in Process Mining and Machine Learning have brought about several approaches for the extraction of an important form of knowledge: the one that discriminates between two classes of temporal event data using temporal logic patterns. In this exploratory paper, we introduce a framework for analyzing and comparing these different approaches. In particular, the framework is used to test three different state-of-the-art approaches, namely binary discovery, Deviance Mining and explanation-based techniques. While the specific results could be affected by the considered implementations, the evaluation framework is general and enables the comparison of any methods for extracting temporal logic knowledge from temporal event data.