Incremental process discovery

• Authors: Marc Solé;Josep Carmona
• Venue: Transactions on Petri Nets and Other Models of Concurrency V
• Year: 2012

Process Discovery techniques provide an automatic shift between a trace or automata model into an event-based one. In particular, the problem of deriving Petri nets from transition systems or languages has many applications, ranging from CAD for VLSI to medical applications, among others. The most popular algorithms to accomplish this task are based on the theory of regions. However, one of the problems of such algorithms is the space requirements: for real-life or industrial instances, some of the region-based algorithms cannot handle in memory the internal representation of the input or the exploration lattice required. In this paper, the incremental derivation of a basis of regions and the later partitioned basis exploration are presented, which allow splitting large inputs in fragments of tractable size. The theory of the paper has been implemented as the new tool dbminer. Experimental results on medium-sized benchmarks show promising reductions in the time required for process discovery when compared to other region-based approaches.