Incremental Mining Of Business Processes Based On Infrequent Behavior

With the rapid development of the information age,business process management has attracted the attention of enterprises.Good business operation processes can ensure the normal operation of the enterprise and increase the production efficiency.Process mining is a technique that extracts information from event logs to discover,monitor,and improve processes.The data growth rate is too fast,and the results of previous mining may no longer adapt to the current data.Therefore,an incremental process mining is proposed.The purpose is to mine the newly added log data based on the existing mining results to avoid the huge workload of re-mining each time from the beginning,and improve the process mining efficiency.Therefore,it has great theoretical and practical application value for the development of business processes.At present,researches mostly find frequent behaviors in business processes,and filter infrequent behaviors as noise directly,which makes business process models lack some behaviors.In order to solve this limitation,this paper proposes a business process incremental mining method based on infrequent behavior.Based on the infrequent behavior in business processes,this paper proposes an effective infrequent behavior analysis method based on entropy,an incremental mining method based on heuristics,an effective low-frequency mining method based on weights,and an incremental mining based on lists.The effectiveness of the proposed algorithm is verified by example analysis.The main work of this paper is as follows:(1)In view of the increase of infrequent logs,existing studies directly determine infrequent behaviors based on the occurrence frequency,ignoring that infrequent behaviors become frequent in the course of business development.This paper proposes an effective infrequent behavior analysis method based on entropy.First,calculate the occurrence probability of the infrequent log and the newly added infrequent log,delete the infrequent log with a larger condition occurrence probability value to obtain the low-frequency log,and then calculate the direct front set rate and the direct back set of activities in the infrequent log.The rate,the entropy value of the activity is compared with a threshold,and it is determined that infrequent behavior is effective(2)Aiming at the problem that the process mining results do not adapt to the current business process,this paper proposes an incremental mining method.Based on the idea of heuristic mining,a dependency tree is established for the direct dependencies of each activity in the infrequent log,the confidence of each activity is calculated,and the newly added logs are replayed to the logs below the threshold in the initial process model to calculate The confidence level of the activities in the log is used to update the dependency tree until the latest dependency tree is obtained.Compared with the initial process,the latest active dependency is found to optimize the process model.Finally,the validity of the proposed method is verified by an example.(3)Aiming at the importance of different activities in business processes,an effective low-frequency mining method based on weights is proposed.Consider not only the frequency of the terms,but also the importance of the terms.First,an initial model is established for the high-frequency log,and the closeness of the behavior between the infrequent log and the initial model is calculated and compared with a threshold to obtain a low-frequency log.The activity is assigned weights according to the importance in the business process,and the effective low-frequency behavior is obtained by comparing the weighted gain with the maximum gain threshold.Finally,the rationality of the proposed method is verified by examples.(4)Aiming at the problem of effective behavior discrimination in infrequent logs,a list-based incremental mining method was proposed.First,the pattern conversion is performed on the infrequent log,and each pattern is sequentially inserted into the pattern table,the item weight table,and the values in the table are updated and arranged.Then,an average weighting table is established,and the value in the average weighting table is compared with the maximum gain threshold.The items with low importance to the process model are obtained,and the infrequent logs containing these items are deleted.The effective infrequent behavior is used to optimize the process model.Finally,the examples prove that the proposed method is feasible.Figure[14]Table[29]Reference[100]...