Research On Process Mining Algorithm Based On Logs

At present, most information systems are driven by explicit process models. Workflow Management System, and ERP, CRM, B2B are all configured with Workflow Models based on tasks. Designing a process model is a complicated time-consuming process, and there are always some discrepancies between the actual workflow and the model we created. In this paper, we present a new method that support the exact workflow design. We call this workflow modeling method as 'Process Mining From Log', which use the data directly related to the events happened during the execution of existing process to support workflow modeling. The log used in this paper is a so called 'Event Log', which consider every task in the process as an atomic event.The paper begin with the introduction of newest development of workflow technology, also the reference model of the WfMC (Workflow Management Coalition). Then the paper make a conclusions on different kinds of modeling method of process, and give the challenges and chances we faced in the workflow modeling fields.Before discussing the mining method , the paper describs some technology and theory related to process mining , including algebra express of log and definition and property of Petri Net and Workflow Net, creating the mapping relationship between Petri Net and Workflow Net.After that, It comes for the paper to discuss the most challenges of this article-non-structural process mining , especially non-free choice. It includes two sections: process mining algorithm and improvement of process mining algorithm. In the first section, a new process mining algorithm based on logs is brought up, and the implement with java for the algorithm is given. The algorithm has two steps. Firstly, according to the sequence of tasks in logs, the connection among tasks of the process is identified, resulting in a structure of the process graph. Secondly, the relations including sequence, choice, parallel and cycle should be confirmed by analyzing the process logs. In the second section, we improve the algorithm to realize the mining of non-free choice structure. The basic idea is that we use the "default exist" idea ,for all tasks that do not adjacent in every track of logs,the possibility that they are linked exists. Then, according to the analysis of logs, we eliminate the tasks that are impossible to be linked, and determine the choice line. Using our mining algorithm ,we can get a reasonable and security of model.In the end, we get enough log through running the log_produce program, and use these log in our new algorithm . Through the analyses of the experiment results ,we learned that our new algorithm has an obvious advantage in getting the reasonable, security and understandable model...