Evaluating Trace Encoding Methods in Process Mining

Encoding methods affect the performance of process mining tasks but little work in literature focused on quantifying their impact. In this paper, we compare 10 different encoding from three families (trace replay and alignment, graph embeddings, word embeddings) using measures to evaluate overlaps feature space, accuracy obtained, computational resources (time) consumed with a classification task. Across hundreds event logs representing four variations five scenarios anomalies, it was possible identify edge2vec method as most accurate effective reducing class overlapping space.