NFmcp @ ECML-PKDD 2020
Invited talk
Fabrizio Maria Maggi:
Abstract:
Process mining is the part of Business Process Management (BPM) that is focused on the analysis of business processes based on process execution logs (event logs). Another central artifact in process mining are process models that are used as inputs and/or outputs of process mining techniques. Two different approaches can be followed for representing process models: procedural or declarative. Procedural process models aim at describing end-to-end processes and allow only for the process behavior that is explicitly specified in the model. However, procedural process mining techniques applied to processes characterized by a high number of different paths and exceptions produce as output process models that are often unreadable (spaghetti-like models). In these cases, it may be a better choice to use declarative process models that model the process as a (small) set of rules that the process should follow. In this way, everything that is not constrained is allowed and several execution paths can be represented in a compact model. In this keynote, we illustrate goals and challenges of rule mining (i.e., the branch of process mining based on declarative process models) by presenting different rule mining techniques and by showing how these techniques can be applied in practice using RuM, the first comprehensive toolset for rule mining.
Abstract:
Process mining is the part of Business Process Management (BPM) that is focused on the analysis of business processes based on process execution logs (event logs). Another central artifact in process mining are process models that are used as inputs and/or outputs of process mining techniques. Two different approaches can be followed for representing process models: procedural or declarative. Procedural process models aim at describing end-to-end processes and allow only for the process behavior that is explicitly specified in the model. However, procedural process mining techniques applied to processes characterized by a high number of different paths and exceptions produce as output process models that are often unreadable (spaghetti-like models). In these cases, it may be a better choice to use declarative process models that model the process as a (small) set of rules that the process should follow. In this way, everything that is not constrained is allowed and several execution paths can be represented in a compact model. In this keynote, we illustrate goals and challenges of rule mining (i.e., the branch of process mining based on declarative process models) by presenting different rule mining techniques and by showing how these techniques can be applied in practice using RuM, the first comprehensive toolset for rule mining.
