Empowering conformance checking using Big Data through horizontal decomposition (Summary)
(Artículo ya publicado)
Conformance Checking is a key technique in the Process Mining paradigm. It allows to detect anomalies and deviations in business processes, being the cornerstone of the diagnose of business processes, promoting the improvement of the quality of these. It becomes especially critical in scenarios in which (i) business processes are highly complex (i.e., containing multiple loops and casuistry), and (ii) there are large amounts of event logs. The diagnose of anomalies in this type of scenarios is challenging, requiring the use of Big Data techniques. In this work, we present a innovative approach to successfully execute Conformance Checking algorithms in Big Data environments. The approach proposes reducing the complexity of business processes through horizontal acyclic decomposition, where end-to-end cuts of the model are obtained, generating partial representations of the model that, taken together, are equivalent to the original complex model. These partial models are less complex to process by conformance checking algorithm. On the other hand, the approach enables the combination of these partial models with different segments of the event logs. In this way, the workload is distributed in a Big Data cluster. Finally, this work proposes a new conformance checking approach based on constraint optimisation problems. The proposal has been tested with five different challenging data sets with highly complex petri nets. The results showed a better performance than traditional standalone algorithms.
Acceso a los detalles haciendo click aquí.