Research On Incremental Algorithms Of Mining Workflow In Batch Processing Aspect

Batch processing in workflow is to integrate workflow activity instancesgenerated from the same class of activities, changing the separately executing patternof traditional multiple workflow instances to one-time executing pattern by mergingthem. Batch processing of workflow can reduce the cost of activity and improve theefficiency. Because the enterprise business process and application is complex anddynamically changed, comprehensively grasping and understanding ways of workflowbatch processing way is as difficult as definition of workflow process. Workflowmining becomes a new branch of research, when data mining technology is introducedto workflow mining. Workflow mining of batch processing is as important asWorkflow mining. Therefore, aiming at the workflow batch processing mining, thispaper makes the following outcomes or contributions:1) Existing research about workflow batch processing mining, didn’t consider theproblem that the “dirty data” will affect the accuracy of the mining results. Aimed atthis problem, this paper cleans the workflow log data by condition functionaldependency, and presents an incremental discovering algorithm for CFDs, which isnamed as CFUP and is based on CFINDER algorithm. The main idea of CFUP reducescan for the original data by the last of mining result, then reduce the running timeand improve the efficiency. The feasibility and validity of CFUP is proved bytheoretical analysis and experimental results.2) Existing research about workflow batch-favor mining, didn’t consider theproblem that workflow log data will incremental update. Aimed at this problem, thispaper presents an incremental mining algorithm for batch-favor dependency, which isnamed as Inc_Mine_BD and is based on Mine_BD algorithm. When the workflow logdata incremental update, Inc_Mine_BD algorithm scans dataset to decide whetherexisting batch dependency is valid, and the new data produces new batch dependency.Theoretical analysis and experimental results proves the feasibility and validity ofInc_Mine_BD.3) To effective evaluate influence of data cleaning in workflow batch mining,this paper combine CFUP algorithm and Inc_Mine_BD algorithm, clean workflow log data by CFDs from CFUP algorithm, then respectively mine in before cleaning andafter cleaning workflow log data by Inc_Mine_BD algorithm. Mining results provethe validity of the CFUP algorithm for improving workflow batch mining quality...