Analysis Of Business Process Anomaly Detection Based On Attribute And Rule Discriminations

The construction of process models helps business processes to be executed systematically according to design expectations.However,the complexity and diversity of enterprise operations lead to numerous and challenging-to-maintain models.Additionally,variations in the execution process are inevitable,resulting in anomalies.To detect fraud,identify malicious activities,and analyze system failures,this study aims to promptly identify process issues,provide timely and efficient feedback,and make informed decisions.This research proposes process model abstraction techniques based on the behavioral profile relationship of Petri nets and log activities,along with consistency analysis.Furthermore,two types of anomaly analysis methods are introduced to address unexpected situations during actual execution.These methods consider the correlation between deviations among attributes and analyze features beyond internal event attributes from a holistic perspective.The main contents of this paper are as follows:(1)For the refund system’s business process at different levels of granularity,an abstract method based on behavioral profiles is proposed to reduce the number of models describing the same business process at various abstraction levels.By inputting the process model,performing activity clustering,constructing activity behavioral profiles,and deriving coarse-grained higher-level models along with their corresponding workflows,a new refund business process model is presented to validate the generalizability of the coarse-grained model.This method addresses the modeling cost maintenance concerns by effectively adapting to the diverse requirements arising from changes in related business processes.(2)To address unexpected changes and conceptual drifts that may be identified as anomalies in practical business processes,the study analyzes the impact of abnormal event attributes on anomalous cases and reduces the possibility of misclassifying unexpected events as abnormal cases.By calculating the likelihood of activity occurrences and constructing a standard likelihood graph,the algorithm proposes an attribute anomaly rate value algorithm that considers the influence of different perspectives on event anomalies.This approach allows for the identification of the degree of abnormality between different attributes during process execution,further enhancing the study of business process anomaly detection based on event attributes.(3)To extend the analysis of multi-perspective process issues and explore multiple dimensions,a research method for analyzing case attributes is proposed.It calculates the distance in behavior using the Directly Follows Graph(DFG)and determines category scores through case attribute discrimination analysis.Subsequently,by combining data operations with case attribute conditions,constraint data rules are extracted and weighted to comprehensively calculate anomaly scores.This approach enables precise and detailed analysis of process information,leading to more effective anomaly detection.The feasibility and effectiveness of the proposed algorithms have been validated in practical cases and demonstrate certain advantages compared to other methods.Figure [16] Table [9] Reference [68]...