الگوریتم ژنتیک هیبریدی و تشکل قوانین برای بهترین تمرین های جریان کاری داده کاوی

Business workflow analysis has become crucial in strategizing how to create competitive edge. Consequently, deriving a series of positively correlated association rules from workflows is essential to identify strong relationships among key business activities. These rules can subsequently, serve as best practices. We have addressed this problem by hybridizing genetic algorithm with association rules. First, we used correlation to replace support-confidence in genetic algorithm to enable dynamic data-driven determination of support and confidence, i.e., use correlation to optimize the derivation of positively correlated association rules. Second, we used correlation as fitness function to support upward closure in association rules (hitherto, association rules support only downward closure). The ability to support upward closure allows derivation of the most specific association rules (business model) from less specific association rules (business meta-model) and generic association rules (reference meta-model). Downward closure allows the opposite. Upward-downward closures allow the manager to drill-down and analyze based on the degree of dependency among business activities. Subsequently, association rules can be used to describe best practices at the model, meta-model and reference meta-model levels with the most general positively dependent association rules as reference meta-model. Experiments are based on an online hotel reservation system.

The process of finding and performing in-depth analysis on patterns/anomalies within the data is increasingly crucial as these patterns describe relationships among important decision variables which can increase the manager’s understanding of the current state. The decision outcome subsequently informs the determination of strategies, which create competitive advantage. Five competitive forces that most organizations have to deal with are “threat of substitute products, the threat of established rivals, and the threat of new entrants, the bargaining power of suppliers and the bargaining power of customers” ( Porter, 1979). Assessment outcomes often result in adjustments to goals and tactics to outwit existing entrants in the market. Continuous assessment means that the above procedures will be carried out during specific time intervals to understand the current status of the organization and to evaluate the effectiveness of the implemented strategy and next possible actions to be performed. This paper is an extension to our previous three works aimed at introducing approaches to assist continuous assessment and support an organization’s management in decision making. The first earlier work highlights the need to increase the effectiveness of workflow management systems (WFMS) to allow reusability of effective and efficient workflow. This means that WFMS should be extended to provide analysis and create a pool of best practices. It can be achieved by creating a pool of best workflow practices within a repository. This is addressed with the introduction of the Weighted and Layered WF evaluation (WaLwFA) methodology ( Lee & Lim, 2007). WaLwFA adopts the concept of Model-Driven Architecture (MDA) with aims to capture best practices, which can be adopted and instantiated to many domains or various information systems. In WaLwFA, the business process models are evaluated using a set of weighted criteria and sub-criteria, which are derived by averaging the assignment of weights by a group of experts. The business models with higher scores are kept within the repository to form business meta-models and reference model respectively. The repository will constantly be updated when there is a new “good” business model, thus ensuring that the repository remains updated with the latest best practice. The second earlier work highlights the need to have an extensive reusable business performance measurement framework that can pinpoint causal relationships between the organization’s current business performance and its future directions as well as measure the organization’s workforces. To address the second problem, Integrated Model-Driven Business Evaluation (IMoBe) methodology is proposed ( Lim & Lee, 2008). In this framework, a model-driven knowledge base serves as repository to store two or more commonly used business performance strategies or elements to evaluate any business organization such as Sun Tzu’s 13 themes of business management strategies ( Lee, Roberts, Lau, & Bhattacharya, 1998) reported in Ko and Lee (2000) as well as concepts of critical success factor (CSF) and critical barrier factor (CBF) ( Niazi, Wilson, & Zowghi, 2005). IMoBe’s methodology is an integration of several business performance approaches consisting of the Balanced Scorecard (BSC) ( Kaplan & Norton, 1992) and the Quality Function Deployment (QFD) methodology. The selected business performance model will serve as a predictor measure and guide organizations on the next course of actions. The selected model contains criteria and they represent the rows in the House of Quality (HOQ) matrix while the columns are represented by strategic objectives identified for each perspective of the BSC. Customization of criteria is allowed where other criteria can be added to the HOQ’s row in order to assist any organization in making efficient and effective actions. The third previous work describes the need to improve existing evaluation methods in the first problem to extend from weighted evaluation to incorporating DSS with OLAP to improve the process of decision making where rules consisting of criteria as attributes can be hierarchically arranged and sorted according to its degree of complexity with each association rule having scores of support, confidence and correlation. For the third problem, a decision support system architecture consisting of our business performance methodology, namely WaLwFA, is extended to incorporate business intelligence capabilities to assist decision-making ( Lim & Lee, 2010). The C4.5 decision tree algorithm ( Quinlan, 1993) is used to discover significant attributes and association rule, namely the Apriori algorithm, ( Agrawal, Imielinski, & Swami, 1993) is used to derive simple and complex association rules as well as to perform correlation analysis to calculate the dependency among attributes. This paper extends from our DSS-OLAP study (Lim & Lee, 2010) to address the problems below

We have enhanced the existing WaLwFA, which currently supports DSS and OLAP during the decision-making process. We have also improved on the DSS_OLAP component in the IMoBe framework. We have introduced GA as hybrid to the existing association rule algorithm to achieve two objectives. The first objective is to complement the existing association rule algorithm to support upward closure through the introduction of correlation to replace support and confidence as measures in association rules and fitness function in GA. The second objective is to introduce correlation as fitness function in GA. This serves to allow optimization and reduce search time and search space in order to get highly positive correlated association rules. It also serves to bypass the difficult challenge faced by users in specifying the support and confidence parameter values. Positively-correlated association rules consisting of business processes as items represent high dependency among business processes. Such association rules can serve as best practices to refine strategic planning models and to refine and enhance the design of decision support systems. Experimental outcomes have confirmed the viability of our approach.