سیاهه های مربوط به داده کاوی رویداد برای حمایت از تخصیص منابع جریان کاری
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|21998||2012||12 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Knowledge-Based Systems, Volume 35, November 2012, Pages 320–331
Currently, workflow technology is widely used to facilitate the business process in enterprise information systems (EIS), and it has the potential to reduce design time, enhance product quality and decrease product cost. However, significant limitations still exist: as an important task in the context of workflow, many present resource allocation (also known as “staff assignment”) operations are still performed manually, which are time-consuming. This paper presents a data mining approach to address the resource allocation problem (RAP) and improve the productivity of workflow resource management. Specifically, an Apriori-like algorithm is used to find the frequent patterns from the event log, and association rules are generated according to predefined resource allocation constraints. Subsequently, a correlation measure named lift is utilized to annotate the negatively correlated resource allocation rules for resource reservation. Finally, the rules are ranked using the confidence measures as resource allocation rules. Comparative experiments are performed using C4.5, SVM, ID3, Naïve Bayes and the presented approach, and the results show that the presented approach is effective in both accuracy and candidate resource recommendations.
Workflow is now an embedded technology in many enterprise information systems (EIS, e.g. PLM, ERP, CRM, SCM and B2B applications). Workflow resource allocation serves as an indispensable link between workflow activities and resources, and it directly determines the execution quality of the workflow activities ,  and . Based on our investigation, most of the resource allocation tasks in present workflow management systems are usually performed using a role-based approach ,  and . That is, to divide the workflow resources (actors) into different candidate groups based on their role and the organization properties. Once the workflow cases are originated, the workflow engine assigns the works to proper resource groups  and . Such resource allocation is somewhat coarse-grained and may fail in some situations. For example, in the manufacturing enterprises, a manufacturing process sheet work might be predefined to be undertaken by the resources with the role “process planning designer”. Actually, some of the processes planning works have to be further assigned to a smaller group of one or more qualified designers instead of all the process planning designers. Thus, the present resource allocation methods may make inappropriate staff assignments and the final quality of the products may suffer from it. Therefore, in some industries such as the manufacturing enterprises, most of run-time workflow resource allocation works are still performed manually by the administrators. The number of administrators is usually small, whereas the activities are of great abundance in some cases. That makes it a time-consuming work to allocate the workflow resources manually. Fortunately enough, contemporary workflow applications usually record the business events in event logs. These logs typically contain information about events referring to a case, an activity, and an originator , ,  and . The case (also referred to as process instance) is a work that is being handled, e.g. a process planning sheet design, a compressor design, an NC programming, etc. As the atomic element of the case, an activity is an instance of a workflow task. An originator is a resource (usually a person) that executes the activity . In this paper, a Process refers to a workflow template of the case, a Task represents a series of similar activities, and a Resource refers to a task performer. This paper presents an Apriori-like algorithm  and  to find frequent patterns from the workflow logs, which are used to generate rules according to a “resource allocation rule constraint”. All the negative correlated rules are annotated with a rule evaluation measure referred to as “correlation measures”. Then, the selected rules are ranked in a descending sequence by their confidence, and the final rules are then recommended to workflow administrators at workflow run-time. The major contributions of this paper are as follows: First, it designs a closed-loop workflow framework for a more intelligent and finer-grained resource management. Second, it proposes an association rule mining approach to find the logics between workflow resources and the activities, which would help decision-making in resource allocation. The remainder of this paper is presented as follows: In Section 2, we design a closed-loop workflow architecture for optimizing resource allocation. Later on, we study the workflow event models and their relationships in Section 3, and then propose our mining approach in Section 4. In Section 5, we empirically compare some classification algorithms (C4.5, SVM, ID3, and Naïve Bayes) with our approach. In Section 6, we discuss some possible improvements. Finally, we discuss the related works in workflow resource allocation in Section 7, and conclude this paper in Section 8.
نتیجه گیری انگلیسی
We have presented a decision-making approach using data mining technology to make recommendations to workflow initiators. In the closed-loop workflow resource allocation framework, the association rules mining algorithms are applied to the workflow event log for mining resource allocation rules. Our current research is oriented towards developing more productive WfMS in resource management along the following lines: (a) implementing the proposed framework in a web-based architecture, (b) association rules mining to generate strong resource allocation rules, (c) using the negative correlation measures to annotate the negative correlated rules, (d) ranking the rules to make decision support for resource allocation. To illustrate, we make some comparison experiments on the log data distracted from a manufacturing enterprise, experiment results show an overall accuracy of over 50%, and we made a comparison between the presented approach and the classification algorithms and analyzed their performances. Feasibility evaluation via a case study suggests that the proposed approach would be useful in supporting workflow resource allocation. Then we discuss the advantages and limitations of the method. Along with the administrators’ awareness of the workload of the resources, and professional knowledge to different product design tasks, our approach can well handle most of the resource allocation problems in PAISs. Our future work includes two main parts: (1) compare some other machine learning approaches like inductive learning programming (ILP) with our present method to find some more efficient and effective approaches. (2) find the resource allocation rules from different organizational levels and dimensions (e.g. the roles and the organizational units).