Business process management plays a pivotal role in enterprise administration.By methodically strategizing,refining,and overseeing business processes,companies can amplify their efficiency and excellence,while mitigating expenses and risks.In this context,process mining constitutes an essential technical instrument in business process management.Process mining can reveal patterns and issues that necessitate attention by scrutinizing and analyzing available business process data,thereby enabling optimization and improvement.This facilitates the efficiency and quality of business processes,reduces redundant manual intervention,and strengthens precision and real-time decision-making.Nevertheless,traditional process model repair methods suffer from some deficiencies,such as a lack of flexibility and adaptability to diverse scenarios,resulting in suboptimal practical outcomes.To counter these issues,process models require repair and enhancement.Model repair is a technical tool in process mining that aims to improve the fitness and precision of existing process models by patching and refining them.The focus of this research is on model repair techniques in process mining,which aim to enhance and repair process models by mining and analyzing existing business process data to improve the efficiency and quality of business process management.The main contents are as follows:(1)Aiming at the problem of generating no new activities in the event log and the inability to replay event logs accurately in the initial model,the current methodology for identifying deviations primarily employs alignment replay techniques.However,this approach lacks a quantitative analysis of the abstraction’s structure from a behavioral perspective.In this paper,we propose a behavioral profile-based control flow deviation detection and model repair method.The proposed technique analyzes log and model deviations using behavior profiles and repairs the original model by logical Petri nets.Firstly,we calculate compliance between the log and model based on the behavioral profile to identify deviation traces.Next,we select logical transitions from the deviation activities based on the set of deviation triples in the deviation traces.Finally,we set the logical function based on the logical transitions to repair the original model by either adding new branches or reconstructing the new structure.(2)Aiming at the problem of generating new activities in the event log and the resulting deviation of the event log from the original model in the control flow perspective,the current detection method lacks consideration for attribute aspects and fails to analyze the event log and the model from the data flow perspective.This paper proposes a method of control flow deviation detection and model repair when data flow is consistent.Firstly,we obtain the base likelihood graph by comparing the actual event logs with the original model.Next,we perform deviation detection using the process tree and the actual event logs,respectively.Finally,we identify problematic nodes based on the detected deviation activity behavior contour relationship and employ the logic function to add constraints to repair the original model by inserting new selection branches between the corresponding variation activities.(3)Aiming at the problem of generating new activities in the event log and deviating from the original model in both control flow and data flow perspectives,this paper proposes a deviation detection and model repair method under data flow and control flow fusion.Firstly,the deviation detection is performed according to the behavior profile and attribute alignment,and when there is a deviation in the data flow but no deviation in the control flow detection,the deviated fragment is repaired by means of segment adaptive;when both data flow and control flow produce deviation,the deviated fragment is constrained,the adapted fragment is inserted into the original model,and the original model is repaired by adding logical variation for behavior constraint according to the structure type.Finally,it has been experimentally verified that the model repair method proposed in this paper can systematically analyze the deviations between models and logs,enhances model quality,maintains the fitness of 1 or within a reasonable range,and effectively improves the precision of the repair model while accurately expressing the semantics of the event log.Figure [18] Figure [54] Reference [67]... |