Detecting Data-model-oriented Anomalies in Parallel Business Process

Abstract

Currently, most information systems are data intensive. The data models of such are posing notable influence on business processes. However, the predominance of existed process verification methods leave out the impact of data models on process models. Meanwhile, with parallel structures in business processes multiplying, business process structures are becoming increasingly intricate and large in size. A parallel structure engenders also uncertainty, and consequently increases the chances and decreases the detectability of anomalies occasioned by process and data model conflicts. In this paper, these anomalies are analyzed and classified. A data state matrix and data operation algebra is introduced to establish the relation between the parallel-process model and the data model. Then, an anomaly detection method under the divide-and-conquer framework is proposed to ensure efficiency in detecting anomalies in business processes. Both theoretical analysis and experimental results prove this method to be highly efficient and effective in detecting data model oriented anomalies.