Efficient edge filtering of directly-follows graphs for process mining

Automated process discovery is a mining operation that takes as input an event log of business and generates diagrammatic representation the process. In this setting, common generated by commercial tools directly-follows graph (DFG). some real-life scenarios, DFG contains hundreds edges, hindering its understandability. To overcome shortcoming, generally offer possibility filtering edges in DFG. We study problem efficiently extracted from while retaining most frequent relations. formalize optimization problem, specifically, finding sound spanning subgraph with minimal number maximal sum edge frequencies. show instance NP-hard outline several polynomial-time heuristics to compute approximate solutions. Finally, we report on evaluation efficiency optimality proposed using 13 logs.