دانلود مقاله ISI انگلیسی شماره 21753
عنوان فارسی مقاله

چارچوب استدلال موردی مبتنی بر مدل مدیریت گردش کار

کد مقاله سال انتشار مقاله انگلیسی ترجمه فارسی تعداد کلمات
21753 2004 29 صفحه PDF سفارش دهید محاسبه نشده
خرید مقاله
پس از پرداخت، فوراً می توانید مقاله را دانلود فرمایید.
عنوان انگلیسی
A case-based reasoning framework for workflow model management
منبع

Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)

Journal : Data & Knowledge Engineering, Volume 50, Issue 1, July 2004, Pages 87–115

کلمات کلیدی
مدل سازی مورد گرا جریان کار - استدلال مبتنی موردی - موقت جریانات کاری - مدل استفاده مجدد
پیش نمایش مقاله
پیش نمایش مقاله چارچوب استدلال موردی مبتنی بر  مدل مدیریت گردش کار

چکیده انگلیسی

In order to support efficient workflow design, recent commercial workflow systems are providing templates of common business processes. These templates, called cases, can be modified individually or collectively into a new workflow to meet the business specification. However, little research has been done on how to manage workflow models, including issues such as model storage, model retrieval, model reuse and assembly. In this paper, we propose a novel framework to support workflow modeling and design by adapting workflow cases from a repository of process models. Our approach to workflow model management is based on a structured workflow lifecycle and leverages recent advances in model management and case-based reasoning techniques. Our contributions include a conceptual model of workflow cases, a similarity flooding algorithm for workflow case retrieval, and a domain-independent AI planning approach to workflow case composition. We illustrate the workflow model management framework with a prototype system called Case-Oriented Design Assistant for Workflow Modeling (CODAW).

مقدمه انگلیسی

Business process modeling is a critical activity in modern organizations to enable enterprise application integration, standardization of business processes, and online B2B and B2C E-commerce. The importance of business process modeling in IT-enabled business process management strategies is indicated by the recent calls for improving process modeling and process management in organizations [14]. Workflow modeling involves the translation of high-level business requirements into workflow schemas that can be executed by appropriate workflow engines. Specifying a workflow model is a knowledge intensive endeavor because development of a typical workflow model requires detailed understanding of the business process logic, the organizational chart, and the information systems accessed by the workflow. Further, a given informal business process description may be modeled in multiple ways, depending on the underlying IT infrastructure and business context. Recent research has suggested the reuse of field tested process knowledge and associated process models to guide workflow modeling and design efforts [18], [20] and [24]. Towards this end, recent commercial systems (such as Oracle Workflow–11i, INCOME [35], and ARIS [39]) provide basic templates for business processes, such as order processing and procurement. These templates may be instantiated and appropriately modified to an organization's needs by a knowledgeable workflow designer. However, standards for template representation and associated ontologies, formal guidelines for reuse of these templates, rules for their instantiation or modification, and procedures for their composition into complex workflows are currently non-existent. Furthermore, there is a lack of design guidelines and workflow modeling tools at present to support tasks such as generation of alternative workflow process models for a given set of business requirements [40]. Workflow modeling and design is the task of defining structured workflow schemas from informal business requirements that satisfy a variety of business logic, organizational and resource constraints. We note that the terms workflow schema, process model, and workflow model are used synonymously in this paper. Workflow modeling involves definition and selection of appropriate tasks (possibly from a task library), sequencing of the tasks to satisfy data and logical dependencies, allocation of resources consumed by the tasks, allocation of agents to execute tasks, scheduling of tasks considering concurrency, and finally, validating and verifying the model [1]. Manual workflow modeling is supported by graphical interfaces, where the workflow model is defined as a graph. Workflow modeling involves searching (albeit implicitly) through a design space defined by a large number of process model alternatives and selection of an optimal process model to fulfill the given problem. With the increasing adoption of workflow management systems and the advent of flexible process integration technologies such as web services, there is an acute need for developing tools and approaches to support workflow design and modeling. In this paper, we propose a workflow model management framework based on a structured workflow design process, which enables reuse of process knowledge (both structured and unstructured) from organizational process repositories. The workflow design process consists of two phases. In the first phase, relevant business tasks are ordered into a seamless whole, satisfying pre-conditions and post-conditions. The result of this phase is a workflow model, a project network defined by a partial ordering amongst all the relevant tasks. Multiple workflow models may be designed to fulfill a given set of business goals. In the second phase, a process model is selected from the available alternatives and is further annotated with appropriate agents, resources and timing information, followed by incorporation of routing details, such as appropriate forks and joins to facilitate concurrent execution. Both phases of design may reuse process knowledge from available repositories. We use a case-based reasoning (CBR) approach [22], which consists of case retrieval, case reuse, case adaptation and case verification tasks, to support workflow model reuse during workflow design. CBR is a computational approach that supports explicit reuse of partial and possibly incomplete, experiential knowledge (stored as cases) in solving ill-structured and complex cognitive tasks, such as design. Past knowledge may be reused to explore the workflow design space and synthesize new solutions. The CBR-approach to workflow model management has been prototyped in a system called Case-Oriented Design Assistant for Workflow Modeling (CODAW), which currently supports the first phase of design mentioned above. Development of an effective case representation, similarity-based retrieval algorithms, and case composition techniques are essential steps in the development of a CBR system to support workflow modeling and design. The main contributions of our research are: • The definition of a structured workflow case representation that includes both declarative and procedural descriptions. The procedural description is based on the process graph metamodel detailed in [3] and the declarative description is based on a predicate logic-based situational calculus formalism of Artificial Intelligence (AI) planning [38]. • The development of a similarity flooding for workflow (SFW) algorithm to support retrieval of procedural workflow models, based on inexact matching of graph queries. The algorithm is derived from the similarity flooding (SF) algorithm [30], recently proposed for metadata model management in database systems. It is based on identifying local structural similarities between two directed labeled graphs, namely, a query graph and the process graph model of a workflow schema. • The development of a workflow composition procedure using the Hierarchical Task Network (HTN) AI planning technique [32] and [34]. CODAW uses the declarative representation (mentioned above) to compose workflow models consisting of both sequential and concurrent tasks that support fulfillment of business requirements encoded as goals involving transformation of an initial business state into a final goal state. The remainder of the paper is organized as follows. In Section 2, we discuss the relevance of case-based reasoning for workflow modeling and the recent advances in model management and AI planning. Section 3 provides an overview of CODAW, including the representation for workflow cases and organization of the process repository. In Section 4, we present the procedures for case retrieval and composition. Discussion of the proposed approach and concluding remarks are given in Section 5.

نتیجه گیری انگلیسی

Development of effective case retrieval and case composition algorithms is essential for the utilization of CODAW for workflow modeling in real world scenarios. The proposed case representation supports similarity-based case retrieval using the process graph formalism and planning-based case composition is facilitated using the declarative representation. Our experience in developing the case repository in CODAW has highlighted the complexity of obtaining workflow cases that may be reusable in an automated manner. Commercial workflow representations are proprietary and their buildtime XML-based representations (if available) are cumbersome to manipulate. Further, workflows created using such tools may use non-standard ontologies for labeling the workflow process models. This has motivated our development of the SFW algorithm to cope with the ontological mismatch problem. The underlying SF algorithm has proven to be reliable in large-scale conventional database schema matching problems. We have conducted extensive simulated experiments to evaluate the robustness of the SFW algorithm to changes in similarity and graph topologies. SFW is sensitive to relative values of initial similarity estimated by the string-matching algorithm. We are currently developing reliable means to assign initial similarities based on other properties (of cases) such as the different structural elements in the case representation such as the task types, inputs, outputs and history. The development of the declarative representations, including the associated predicates and state descriptors, to support case composition is a complex task. A process graph does not provide explicit insight into the underlying design intent of why a particular task was chosen and instantiated in a given schema. It does not support reasoning about the interaction between the tasks. In contrast, a plan is a process model with state, developed from explicit encoding of the domain knowledge, which facilitates reasoning about choices of tasks and their inter-relationships. Developing appropriate declarative representations requires background domain knowledge and acquiring such knowledge is difficult. Development of such declarative representations is essential for the convergence of the Semantic web and workflow technologies in the short term. We note that our choice of first order predicate logic representation for case composition has been guided by its use in real-world planning systems and formal models of Semantic Web ontologies. Currently in CODAW, the SHOP algorithm generates a sequential workflow, which is then post-processed into a workflow with concurrent tasks by analyzing data dependency constraints. This staged approach has been adopted to identify if a solution to a new problem exists, in a timely manner. Consideration of concurrency explicitly during the planning process increases the size of the search space and requires models for temporal and resource reasoning. We are currently experimenting with partial-order planners, that consider concurrency, to compose cases [38]. Our research on planning-based process composition techniques has been successfully used for dynamic web services composition [26]. The declarative and procedural instance-level case representations may be used to infer generic process models using learning techniques such as Inductive Logic programming and statistical machine learning [38]. This paper has described a case-based reasoning approach to support workflow modeling and design. The innovative features proposed in the paper are: (1) a case representation for workflow schemas and instances that combines both declarative and procedural representations, (2) a similarity based retrieval algorithm for retrieving process graphs of workflow schemas based on graph-based queries, and (3) the use of a domain independent AI planning technique to facilitate composition of cases into a workflow. Our future work will continue the development and testing of the CODAW framework. We intend to further develop the adaptation and verification modules for CODAW, provide techniques for retrieval based on events, and integrate the different phases of the CBR design cycle. Enabling case composition using partial-order planners for developing workflow models with concurrent tasks is an interesting and challenging problem. Furthermore, we plan to perform real world user studies comparing the CODAW prototype with commercial workflow tools such as Oracle Workflow–11i.

خرید مقاله
پس از پرداخت، فوراً می توانید مقاله را دانلود فرمایید.