A multi-perspective approach for the analysis of complex business processes behavior

摘要: Abstract Business processes are often monitored by transactional information systems that produce massive dataset called event logs. Such logs contain the process execution traces, typically characterized heterogeneous and high-dimensional data. Process mining techniques offer a great opportunity to gain valuable knowledge hidden in data be used for analysing multiple characteristics of (i.e. perspectives mining, like structural aspects, activities, resources, time). Therefore, raw must encoded into suitable format can more conveniently provided algorithms. However, most existing encoding focus on control-flow perspective, i.e. only encode sequence activities characterize trace, leaving out other fundamental describing behavior all its aspects. In this paper we address problem computing concise informative representation traces considers behavior. We propose holistic approach computes trace embedding able capture patterns dependencies between lost one-dimensional analysis and, at same time, it is unsupervised, meaning no priori needed. The experiments conducted two real life demonstrate our proposed appropriate concisely describe various processes, method outperforms techniques. Furthermore, includes elapsed time events as an additional feature make us capable use further dimension analysis.