Advanced Process Discovery Techniques

Abstract Given the challenges associated to process discovery task, more than a hundred research studies addressed problem over past two decades. Despite richness of proposals, many state-of-the-art automated techniques, especially oldest ones, struggle systematically discover accurate and simple models. In general, when behavior recorded in input event log is (e.g., exhibiting little parallelism, repetitions, or inclusive choices) noise free, some basic algorithms such as alpha miner can output However, complexity data increases, quality discovered models worsen quickly. that oftentimes real-life logs record very complex unstructured containing infrequent traces, incomplete data, techniques turn unreliable not purposeful. Specifically, they tend either have limited accuracy (i.e., low fitness and/or precision) are syntactically incorrect. While currently there exists no perfect technique, better others discovering model from recording behavior. this chapter, we introduce four discussing their underlying approach algorithmic ideas, reporting benefits limitation, comparing performance with introduced previous chapter.