Attributbasierte Ressourcenzuweisung in Geschäftsprozessen

Abstract

In Business Processes, activities are frequently executed using both physical (e.g., machines, employees) or virtual resources. Most resource allocation procedures in BPMS choose randomly a resource which satisfies a given role. In a company a role can be occupied by more then one resource, resulting that a selection-algorithm must be used to decide which resource executes a given activity. Choosing the optimal resource is not always easy and the best possible selection can be achieved if all resource attributes like education, training data and past execution data are factored in. If not all of the data is factored in, the probability of not finding the optimal resource is increasing and suboptimal results are achieved. This student research project presents an approach for the optimization of resource allocation in business processes. It explores both the concepts and their prototypical implementation. Business Processes are composed of business activities which can be executed parallel or sequential. They receive input data and turn it into output data which should bring benefit to the customer, supplier or the company itself. Input data in the student research project are resources. Resources are implemented as staff and machines. The optimal resource-allocation of staff or machines along multiple dimensions like time, costs or success rates can positive influence business processes. This student research project implements a resource manager which finds the best available resource, under consideration of several optimization parameters, in a company. To achieve this goal, several resource allocation procedures are analysed and a model-based procedure is presented which builds regression trees from the resource attributes. The regression trees can predict which resource will perform best a given activity. One algorithm which can build regression trees and is also being used in the student research project is the M5P algorithm. The M5P algorithm has the advantage over the normal regression being able to handle non linear data. On the basis of a typical business process where resources are freely distributed, several optimization scenarios are analysed and compared how the model-based algorithm behaves compared to the standard algorithm in BPMS. At the end, expandability possibilities are discussed which can further improve the quality and the effectiveness of the resource manager.