An Information-Theoretic Framework for Process Structure and Data Mining

Process-oriented systems have been increasingly attracting data mining community, due to the opportunities the application of inductive process mining techniques to log data can open to both the analysis of complex processes and the design of new process models. Currently, these techniques focus on structural aspects of the process and disregard data that are kept by many real systems, such as information about activity executors, parameter values, and time-stamps. In this paper, an enhanced process mining approach is presented, where different process variants (use cases) can be discovered by clustering log traces, based on both structural aspects and performance measures. To this aim, an information-theoretic framework is used, where the structural information as well as performance measures are represented by a proper domain, which is correlated to the "central domain" of logged process instances. Then, the clustering of log traces is performed synergically with that of the correlated domains. Eventually, each cluster is equipped with a specific model, so providing the analyst with a compact and handy description of the execution paths characterizing each process variant.