A Fuzzy-based Approach to the Management of Agile Processes.
Oliver Thomas, Otmar Adam, Christian Seel · 2005
The recent dynamics in business and the rough environment of globalization impose enormous pressure on every enterprise to be innovative. Consequently, management concepts and IT systems have to support adaptive organizations with agile processes. These processes can react to changes in their environment automatically or provide mechanisms to be adapted. A fuzzy based approach enabling the management of such agile processes is presented. 1 Vagueness in Business Process Management Until now only standardized processes have been suitable for Business Process Management (BPM) systems. Comprehensive software products which are able to support thousands of transactions a day are available, e. g. the standardized processing of customer orders in retail enterprises or the granting of loans in a bank. In practice several types of business processes depend on highly dynamic knowledge and require decisions based on rapidly changing circumstances. Such decisions are based on vague or qualitative information. In this paper vagueness is understood as the uncertainty in the environmental context of business processes regarding pieces of data and their interdependences. The complexity of the environment and natural language descriptions of real world facts contain informational and linguistic fuzziness. Both the creation of linguistic models and the context sensitivity of linguistic statements contribute to this vagueness. This complexity isn‘t appropriately supported by BPM so far. This weakness derives from the difficulties of translating the ambiguity of human reactions to changes in their environment into new business process models, and of adapting existing models. Thus the transformation of real-world uncertainty, dynamics, and vagueness into crisp process models based on propositional logic implies a loss of relevant information. The use of vague data is favorable whenever the environment is characterized by high dynamics or is not determinable at build-time. In 1965 Lotfi A. Zadeh proposed an extension of the traditional crisp set theory [9] which allows not only true or false as logical values but also so-called membership values between 0 and 1 and is therefore able to tolerate a certain amount of vagueness.[3][6] With fuzzy sets linguistic variables can be formulated and fuzzy logic controllers can be implemented. The outstanding advantage of these systems is the sophisticated control mechanism, which is achieved using only a small set of very simple rules.[10] The decision rules aren‘t fixed at build time when the process models are designed but are interpretable at run time according to natural language guidelines representing the knowledge behind agile processes. In the field of enterprise modeling some traditional methods have already been extended by fuzzy set theory concepts. Zvieli and Chen[11] describe a fuzzy extension of the Entity Relationship Model (ERM)[4]. Production Planning and Control can particularly benefit from provisional data that represents preliminary information or estimations about customer orders.[1] Petri Nets have been extended by fuzzy concepts in order to represent procedures with vague conditions or incomplete information. Becker, Rehfeldt and Turowski[2] introduce provisional data into Event-driven Process Chains (EPC) as shaded elements. 2 Fuzzy-extended adaption of process models Thomas, Husselmann and Adam[8] have extended the EPC semantically by fuzzy constructs. To achieve these extensions, entities able to represent linguistic variables and fuzzy rule sets have been added to the meta-model of the EPC. This allows to model adaptive processes,[5] as illustrated in the following example: An enterprise has to decide whether a customer‘s order will be accepted and processed or rejected. After the customer has called to place his order, four criteria must be taken into consideration: the technical feasibility, the economic feasibility, the customer‘s credit rating and the product availability. The order will only be accepted if all four criteria are fulfilled. In all other cases the order will be rejected. In fig. 1 the corresponding crisp business process model is shown. Customer’s order defined Check economic feasibility of order Check credit rating of customer Check product availability Check technical feasibility of order Order is technically feasible Order is not technically feasible Order is economically feasible Order is not economically feasible Good credit rating Poor credit rating Product available Product not available Accept customer’s order Reject customer’s order Customer’s order is accepted Customer’s order is rejected Fig. 1. Crisp EPC of customer order process Traditionally the vague conditions would be translated into crisp decision rules leading to rigid process models (assuming that this process were supported by BPM at all). The decision intervals would be limited by exact boundaries and a more or less comprehensive model would substitute the experience of the employee. Relevant information would be lost and the IT supported decision would be inferior to the original one. Introducing fuzzy-extension of EPCs improves business process management. Decision rules are now implemented with the aid of linguistic variables and natural language rule sets. The control logic formalism is taken out of the model so that only the business knowledge is represented in the EPC. This means that any employee can use it directly without major efforts in training. The necessary control knowledge to transfer the business model into workflow controls is attributed. The models can thereby be easily adapted for similar cases or used instantly as reference models expressing the best local practice.[7] The natural way of human thinking is directly applicable in the design and administration of model-based workflow systems. The user can change the easily understandable rule sets at run-time and is therefore able to influence the execution of the business process in a straightforward manner. In fig. 2 a prototype for a business process management tool is shown in a screenshot. A standard BPM tool can be extended to include fuzzy components. The user can work in easy mode where only the traditional left hand side window, containing business content, is displayed. In the right hand corner additional components are available to define or fine tune membership functions using a wizard function and to intuitively enter the rule base. Repeated use of existing standards and user support through wizards ensure the acceptance of the new concepts in practice. Fig. 2. Screenshot of fuzzy modeling tool