Research On Process Mining Based On Genetic Algorithm

Workflow are a reflection of the computerized model, are at an advanced computer environment in order to support the implementation of business process integration and business process automation and the set up by the workflow management system to implement the business model. The workflow life cycle consists of four phases: workflow design, workflow configuration, workflow execution and workflow diagnosis. The process mining is not just a tool of workflow design, but it is very useful for understanding the current business process. The goal of process mining is to reverse the process and collect data at runtime to support workflow design and analysis.This paper firstly introduced the latest workflow technology development, as well as the WfMC workflow reference model, and then it summarizes the main workflow modeling method, opportunities and challenges existing in the field of workflow modeling. Before discussing the mining method, the paper describes 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. Paper then describes the current process mining field of relatively perfectα-algorithm, pointing out shortcomings and limitations ofα-algorithm on the mining of some structures.Due to the deficiencies and drawback of the existing process mining algorithm, as well as the characteristics of genetic method including self adaptive, global optimization, the implicit parallelism and the form of easy to understand, we introduce genetic method to process mining. Before the introduction of the genetic process mining, we define: internal representation, fitness measure and genetic operator. The internal representation defines the search space of a genetic algorithm. The internal representation that we define supports all the problematic constructs, except for duplicate tasks. The fitness measure evaluates the quality of a point (individual or process model) in the search space against the event log. Genetic operators ensure that all points in the search space defined by the internal representation may be reached when the genetic algorithm runs. The above three definitions prepare for the further introduction of genetic process mining.Finally we propose the process mining based on genetic algorithm. This algorithm starts with an initial population of individuals. Every individual is assigned a fitness measure to indicate its quality. An individual is a possible process model and the fitness is function that evaluates how well an individual is able to reproduce the behavior in the log. Populations evolve by selecting the fittest individuals and generating new individuals using genetic operator.In the end, we get sufficient log through running the log produce program, and use these logs in our new algorithm. Through the analyses of the experiment results, we learned that our new algorithm has obvious advantage comparing withα-algorithm in the mining of some structures.