| With the rapid development of computer technology,business process management systems are also widely used.Business processes are a series of structured activities performed to perform tasks,integrating a wide range of data sources,applications,IT systems,and managing enterprise information.Business process management system executions generate large amounts of data,and logs can contain many events,each with different attributes,making it a tedious task to check and analyze the various executions of different business processes.Process mining is a research area dedicated to the a posteriori analysis of business process execution,process mining allows analysts to use the historical execution logs of business processes to gain insight into the actual performance of those processes.As one of the most important tools for modern business process modeling and analysis,Petri net not only provides a clear and simple representation of the structure of the system,but also enables a more accurate simulation of the system's operating processes.This master's thesis investigates the extraction and analysis of business process anomalous behavior based on the relevant knowledge and theoretical nature of the Petri net.There are complex dependencies between the activities of a business process,and operations will produce behavior that does not conform to the dependencies,both process anomalies.Businesses are interested in anomalous behavior in business processes because these anomalies may be indicators of identifying inefficient employees or fraudulent activity,and detecting potential business process anomalies can greatly help business analysts identify and understand the causes of process errors,thereby improving the economic efficiency of the business.There are many current approaches to business process anomaly analysis,and researchers have done a lot of work on business process anomaly analysis.Detecting anomalies in the behaviour of processes at runtime is critical:they may reflect security breaches and fraudulent behaviour on the one hand,and deviations that are required due to,for example,abnormal conditions on the other.In this master's thesis and from the point of view of combining control flow with data flow,a new approach is proposed to analyze anomalies in business processes.The main contributions of this master's thesis are as follows.(1)First,based on the applicability of the modeling of Petri net,the business processes in the implementation of ERP were improved to address the problems of inventory management in the construction of communication cables.The modeling system is optimized by adding the relevant control structure and successfully avoiding the delay in the construction of the communication fiber optic cable due to insufficient materials.One of the core of business process management is the optimization of the business process model,which will affect the management effectiveness and economic efficiency of the enterprise.The experimental results show that the optimization model is reasonably well bounded and safe.(2)The event logs is the starting point of process mining,the current process mining field focuses on mining the frequent behavior in the log,but the infrequent behavior of the log may also have important implications for the behavior management of business processes,the current process mining method is to ignore the infrequent behavior of the logs.In order to extract infrequent behaviors from the event logs and repair them,an event data repair method is proposed in Chapter 4 of this master's thesis:First,the event logs in the system are used to generate the log automaton,then the low-frequency behaviors in the logs are extracted and the infrequent behaviors in the logs are removed to get the anomaly-free automaton,then the log automaton filters the event logs,then the filtered logs are repaired by overriding the probability formula,and finally the repaired logs are shown to have a better fit to the process model by a case.(3)If anomalies in a business process are analyzed only from the control flow perspective,there will be many class rows of anomalies that cannot be identified,and a two-pronged approach to anomaly detection from both the control flow and data perspectives is presented in Chapter 5 of this master's thesis:A probabilistic automaton with roles is trained by combining active executors in the system model with control flows.By mapping the cases that perform anomaly detection to the role probability automaton,it is possible to detect anomalous behavior that cannot be judged from the control flow perspective,and by calculating the probability of occurrence of the current detection process trace,it is possible to locate its anomalous point and thus infer the cause of its occurrence.If the case cannot be mapped to the role probability automaton,or if the probability is less than a given threshold,the case is marked as an exception.Figure[16]table[9]reference[85]... |
