Towards process instances building for spaghetti processes

Process Mining techniques aim at building a process model starting from an event log generated during the execution of the pro- cess. Classical process mining approaches have problems when dealing with Spaghetti Processes, i.e. processes with little or no structure, since they obtain very chaotic models. As a remedy, in previous works we pro- posed a methodology aimed at supporting the analysis of a spaghetti process by means of its most relevant subprocesses. Such approach ex- ploits graph-mining techniques, thus requiring to reconstruct the set of process instances starting from the sequential traces stored in the event log. In the present work, we discuss the main problems related to process instances building in spaghetti contexts, and introduce a proposal for ex- tending a process instance building technique to address such issues.