داده کاوی فرا سازمانی جریان کاری مشارکتی از ورود به منبع چندگانه

Today's enterprise business processes become increasingly complex given that they are often executed by geographically dispersed partners or different organizations. Designing and modeling such a cross-organizational workflow is a complicated, time-consuming process and requires that a designer has extensive experience. Workflow logs captured by different cross-organizational systems provide a very valuable source of information on how business processes are executed in reality and thus can be used to derive workflow models through process mining. In this paper, we investigate the application of process mining for workflow integration based on the concept of RM_WF_Net, a type of Petri net extended with resource and message factors. Four coordination patterns are defined for workflow integration. A process mining approach is presented to discover the coordination patterns between different organizations and the workflow models in different organizations from the running logs containing the information about resource allocation. A process integration approach is then presented to obtain the model for a cross-organizational workflow based on the model mined for each organization and the coordination patterns between different organizations.

With the development of the Internet and distributed computing technologies, today's enterprise business processes become increasingly complex given that they are often executed by geographically dispersed partners or different organizations. For example, a typical multi-modal transportation business process may require the coordination of several business partners including the sender, the consignor, the carrier, the shipper, the buyer and other related partners. Designing and modeling such a cross-organizational workflow require a workflow designer to have lengthy discussions with the involved workers and managers of different organizations. Therefore, creating a workflow design is a complicated and time-consuming process and there are typically discrepancies between the actual workflow processes and the processes as perceived by the management [21]. Many enterprise information systems, such as enterprise resource planning (ERP), customer relationship management systems (CRM), and workflow management systems (WfMS), usually record information of the execution of business processes in their event logs [21]. Thus, process mining, also referred to as workflow mining, has recently been proposed to distil a structured process description from the event logs of those systems [9]. The goal of process mining is to analyze event logs so as to construct a model that best describes all the recorded instances of a workflow. Recently, cross-organizational process mining has received more attention [18]. Although many papers on process mining have been published [3], [5], [6], [20] and [21], it is difficult to apply existing approaches directly to discover a model for a cross-organizational workflow. This can be attributed to the fact that the running log of a cross-organizational workflow is usually distributed on different servers owned by different partners or different organizations, while most existing process mining approaches assume that a single server is used to manage all the operations of a process [3], [5], [6], [20] and [21]. To address this problem, a process mining based integration approach is proposed in this paper to obtain cross-organizational workflow models. Unlike the existing studies on process mining [3], [5], [6], [20] and [21], the event logs in this study are distributed on different servers located in different organizations. The event logs contain information about resource allocation and messages exchanged, which are two important coordination mechanisms between organizations. The concept of RM_WF_Net is proposed to represent the mined workflows. An RM_WF_Net is a type of extended Petri net that allows representation of resource allocation and messages exchanged in workflows. On the basis of the RM_WF_Net model, four different coordination patterns are defined for cross-organizational workflow integration. A process mining approach is presented to discover workflow models in different organizations from the event logs containing the information on resource allocation. To discover the coordination patterns between different organizations, a middleware has been implemented to integrate the workflow running logs between different organizations. Based on the logs integrated, the coordination patterns between different organizations can be obtained by using the process mining approach. According to the workflow models in different organizations and the coordination patterns between different organizations, a process integration approach is presented to obtain the cross-organizational workflow models. Our approach can help an organization to identify from its own perspective the overall workflow model reflecting how various organizations collaborate. Moreover, when the event log data of all relevant workflows are available, our method can help derive a complete overview of a workflow that crosses multiple organizational entities. In the paper, a multi-modal transportation business process is studied as an example to validate the proposed approaches. However, it is only possible to obtain the overall workflow model from the individual perspective of each organization involved in the multi-modal transportation business process, and to understand the collaboration between the organization's private workflow and other workflows. In many tightly coupled application cases, the approaches proposed in this paper can be used to obtain the overall model for a cross-organizational collaborative workflow; consequently, an example of hospital workflow model integration is used to demonstrate the approaches proposed in the paper. The remainder of the paper is organized as follows. Section 2 presents a review of the related work. Section 3 gives a framework for cross-organizational workflow mining and integration. An example of a cross-organizational workflow used to validate the proposed approach in the paper is also given in Section 3. Section 4 defines the RM_WF_Net model for a cross-organizational workflow and the coordination patterns for different organizations. Section 5 presents the workflow mining for one single organization. Section 6 presents the cross-organizational workflow integration using process mining. Section 7 presents an experiment evaluation for the proposed approach. Section 8 concludes the paper with our contributions and future research directions.

A complex workflow is usually executed across several organizations, and it is not easy to obtain the overall model for a cross-organizational workflow. However, the running log of the workflow systems contains much information about their activities and the relations between activities. We present an approach to applying process mining to discover the model for a cross-organizational workflow from the distributed running log collected from different servers located in different organizations. Because the running log contains information about resource allocation, an RM_WF_Net model is proposed to represent the workflow discovered. In contrast to the traditional considerations of process mining, the objective is not to discover the RM_WF_Net model. We pay more attention to model integration for a cross-organizational workflow based on the models mined. This paper makes two main contributions. • A process mining approach to discovering the coordination patterns between different organizations and the workflow model of each organization for a cross-organizational workflow from the distributed running log that contains information about resource allocation. The mining result is represented in the formalized form of an RM_WF_Net, which can be used to process integration for the cross-organizational workflow. • A process integration approach for a cross-organizational workflow according to the coordination patterns and the RM_WF_Net of each organization. Although the workflow mining approach for one single organization cannot address some special structures such as choices or invisible tasks, there are few application limitations for the approach proposed. Theoretically, the proposed approach can be useful if the workflow model for one single organization is without special structures and the coordination relations between different organizations can be expressed by the given four kinds of patterns, i.e., coordination with synchronized activities, coordination with messages exchanged, coordination with shared resources, and coordination with abstract procedures. By experiments, we find that all kinds of cross-organization workflows can be discovered and integrated by the approach proposed if they can be created and simulated by Bonita. In this paper, we assume that the running log of the workflow used for process mining does not have data noise. However, noisy data may occur when, for instance, the wrong activity is implemented such as one of another activity's pre-activities or post-activities. Obviously, the model mined from the noisy data cannot reflect the true structure and behavior of the workflow. In a cross-organizational workflow system, the running log may contain more noisy data, and in future work we will focus on a detection approach for noise in the running log. Meanwhile, resource allocation among different partners within a cross-organizational workflow is important for avoiding resource conflicts. An integration approach using process mining is presented in this paper to obtain the model for a cross-organizational workflow. How to apply the model, integrated to verify and check resource conflicts within a cross-organizational workflow, will also be addressed in our future work.