Abstract:

Interacting with modern business application systems such as SAP R/3 users leave traces in terms of log-, protocol- and trace data. These data can be extracted and used as a main data source for a user behavior analysis. The main point of a user behavior analysis is to discover and analyze interesting patterns from these data. Application Usage Mining (AUM) uses the Mining Methods to discover and evaluate these patterns. The results of the user behavior analysis should support and simplify the personalization of the system from the users point of view as well as [1,2] enable the development of an As-Is Model as a basis to improve the Workflows of the Business Process of an enterprise [3]. The User Trace File of an ERP system contains data, which describe the information of the accomplished transactions and/or transaction steps. These data provide information about the tasks processed by the user. As an ERP system like SAP/R3 has an implicit Workflow model in contrary to WFMS, the User Trace File of the SAP/R3 logs no allocation of users interactions to the processed Workflow instances. Thus, important information about the Workflow instances is missing, which are necessary for modelling a Workflow model. This information must be extracted therefore first from the trace file. This paper treats the Workflow instance problem and develops algorithms to solve this problem.