Study On Mining Algorithms Of Batch Processing Patterns Based On Functional Dependency

Batch processing is common in manufacturing job scheduling and business processes,and it often presents with an activity in business process or production scheduling process. Using batch processing can save resources and improve efficiency. However, due to constraints of personal competency process modeler, in practice, a number of activities which can be batch processed are not discovered. This may lead to the advantage of batch processing not fully applied. Workflow as a powerful tool for modeling business process has been the concern of both industry and academia. At present, supporting batch processing in workflow systems has raised concerns and has undertaken exploratory research. But there is still much work needing further developing and improving. For example, with the advancement of information technology, more and more business processes are managed, executed and monitored by computer systems, this leads to the availability of a large amount of business data and logs, in which the trail of spontaneously performed batch processing will be recorded. So, it is meaningful to determine which activities are suitable for batch processing according to the business data and logs, to guide the modeling of batch workflow.In view of the above problems, according to the database theory of functional dependencies and mining mothedology, we introduce batch-favoring dependencies to identify batch processing activity. At first, we propose an algorithm―Mine-BD which mines batch-favoring dependencies from the workflow system logs or business executive data without noise and interference. The algorithm takes into account the existence of two cases of the mining data, noise-free data and noisy data. For the weakness that the Mine-BD algorithm can not get batch-favoring dependencies or the gotten ones didn't recorded activities that suits for batch processing in the case of containing noisy data, we propose an optimization method. Using ideas developed for the mining of functional dependencies in database, we introduce the concept of approximate batch-favor dependencies, the approximate batch-favor dependencies record the activities that are suitable for batch processing as well, and it gives the algorithm Mine-ABD which mines approximate batch-favor dependencies from workflow system logs or business executive data.Finally a large number of simulation experiments are conducted to verify the correctness of the algorithm. It proves that the research achievement have practical and theoretical values on efficiency and resource optimization in production or commercial activities, and further it improves the theoretical system of workflow.