Business Process Management Software
Business Process Management (BPM) software is one of the fastest growing segments of the enterprise software market, due to the key role it plays in corporate re-engineering. Using BPM software tools, business people can document the workflow and processes in their enterprises, identify bottlenecks and other impediments to efficiency, and suggest alternative and improved business processes. MERL has been supported the development of Mitusbishi Electric's BPM suite by implementing a discrete-event simulator for process analysis in FY05, and inventing a novel process mining algorithm in FY06.
Background & Objective: The purpose of BPM software is to support the documentation, analysis, monitoring, and re-design of the business processes in an enterprise. MERL has participated in the implementation of Mitusbishi Electric BPM suite with two modules: a discrete-event simulator for interactive analysis of business processes (FY05), and a process mining module that builds a model of a business process in standard notation from execution logs of that process (FY06).
Technical Discussion: To a large degree, the two modules are complementary: while the simulator produces synthetic execution logs from a given BP model, the process mining software produces a synthetic BP model given actual execution logs. In spite of this complementarity, however, the technology and technical challenges behind the two modules are completely different. The discrete event simulator uses efficient random number generation and management of event sequences in priority queues, while the main challenge in process mining is to select the most appropriate BP model among those that successfully explain the execution log. Since there are exponentially many such models, the computational complexity of existing algorithms has been very high. In contrast, MERL's novel algorithm for process mining uses a structured representation of business processes called Workflow Trees, and has complexity only cubic in the number of tasks in the model. To our knowledge, this is the only existing process mining algorithm with (low) polynomial complexity.
Future Direction: The project has completed, and we are working on bringing the two modules to existing customers. The process simulation module has been extended by I-Shi-bu to handle time-varying processes, and is currently being deployed at the call center of one Mitsubishi Electric business unit. The process mining algorithm is undergoing comparison with other mining algorithms within the ProM framework from TU-Eindhoven.
Contact: Daniel Nikovski
Publications:
Nikovski, D.; Kulev, V., "Induction of Compact Decision Trees for Personalized Recommendation", ACM Symposium on Applied Computing (SAC), April 2006 (SAC 2006, TR2006-036)
| Technical Reports: | |
| Workflow Trees for Representation and Mining of Implicitly Concurrent Business Processes | |
Technology Area: Sensor and Data Systems
Modification Date: June 13, 2008
