Business process management systems (Smith and Fingar 2003) provide a fundamental infrastructure to define and manage business processes and workflows. These systems are often called process aware information systems (Dumas, Aalst et al. 2005) since they coordinate the automation of interconnected tasks. Well-known systems include Tibco, WebSphere MQ Workflow, FileNet, COSA, etc. Other types of systems, such as ERP, CRM, SCM, and B2B, are also driven by explicit process models and are configured on the basis of a workflow model specifying the order in which tasks need to be executed. When process models or workflows are executed, the underlying management system generates data describing the activities being carried out which is stored in a log file. This log of data can be used to discover and extract knowledge about the execution and structure of processes. The goal of process mining is to extract information about processes from logs. When observing recent developments with respect to process aware information systems (Dumas, Aalst et al. 2005) three trends can be identified. First of all, workflow technology is being embedded in service oriented architectures. Second, there is a trend towards providing more flexibility. It is obvious that in the end business processes interface with people. Traditional workflow solutions expect the people to adapt to the system. However, it is clear that in many situations this is not acceptable. Therefore, systems are becoming more flexible and adaptable. The third trend is the omnipresence of event logs in today’s systems. Current systems ranging from cross-organizational systems to embedded systems provide detailed event logs. In a service oriented architecture events can be monitored in various ways. Moreover, physical devices start to record events. Already today many professional systems (X-ray machines, wafer stepper, high-end copiers, etc.) are connected to the internet. For example, Philips Medical Systems is able to monitor all events taking place in their X-ray machines. The three trends mentioned above are important enablers for path mining and process mining. The abundance of recorded events in structured format is an important enabler for the analysis of run-time behavior. Moreover, the desire to be flexible and adaptable also triggers the need for monitoring. If processes are not enforced by some system, it is relevant to find out what is actually happening, e.g., how frequently do people deviate from the default procedure.
Path mining can be seen as a tool in the context of Business Process Intelligence (BPI). This approach to path mining uses generic mining tools to extract implicit rules that govern the path of tasks followed during the execution of a process. Generally, the realization of a process can be carried out by executing a subset of tasks. Path mining is fundamentally about identifying the subset of tasks that will be potentially be triggered during the realization of a process. Path mining is important to process Quality of Service (QoS) prediction algorithms (Cardoso, Miller et al. 2004). In processes for e-commerce, suppliers and customers define a contract between the two parties, specifying QoS items such as products or services to be delivered, deadlines, quality of products, and cost of services. A process, which typically has a graph-like representation, includes a number of linearly independent control paths (i.e. paths that are executed in parallel). Depending on the path followed during the execution of a process, the QoS may substantially be different. If we can predict with a certain degree of confidence the path that will be followed at runtime, we can significantly increase the precision of QoS estimation algorithms for processes.