Research On Approximate Consistency Analysis And Prediction Method Based On Log Behavior

In the current era of informatization and digitalization,the complexity of business processes is increasing,and the scale of event logs recorded by information systems continues to expand,which leads to the continuous evolution of business processes over time.Therefore,the research on consistency checking has important theoretical significance and application value.It can detect the deviation between the actual process behavior and the expected behavior.However,the traditional consistency analysis methods are difficult to apply to large event logs and process models.In some applications,it is not necessary to calculate the exact consistency between each log behavior and the model behavior.The approximate solution is sufficient to analyze the business process and system.The prediction of business process can predict the possible future behavior according to the historical process behavior,and help enterprises formulate solutions in advance for the system risks perceived in advance,so as to avoid losses to enterprises.Under this background,this paper studies the approximate consistency analysis and prediction methods based on log behavior.The existing consistency analysis and optimization methods tend to focus on the process,behavior,data and other technical aspects of the system,lack of analysis of top-level business objectives,and fail to take more account of the wishes,motivations,strategies and other factors of stakeholders.At present,some related researches have used behavior subsets to solve the approximate consistency of logs and models,but the constructed behavior subsets can not well represent the complete process model,and the accuracy of approximate consistency results is not high.At the same time,in the aspect of business process prediction,the previous methods seldom consider the weak order relationship between log activities.The introduction of behavior profile can accurately describe the interaction and constraint relationship between behaviors,so as to analyze the behavior association relationship between historical activities,which is conducive to predicting future activities.Based on this,this paper has carried out the following work:(1)Aiming at the problem that the existing consistency analysis and optimization methods lack consideration of target based mining,a consistency analysis and optimization method based on target sequence is proposed.Firstly,the target sequence is generated according to the expected behavior of the business process,and the consistency is analyzed and the deviation is found by aligning the target sequence with the process model;Then,the behavior profile relationship between transition is analyzed,and the configuration information is introduced to optimize the process with deviation;Finally,combined with the self-service delivery business process of intelligent express cabinet,the proposed method is verified.(2)Aiming at the problem that the behavior subsets constructed by the existing approximate consistency analysis methods are lack of representativeness,an approximate consistency method based on log clustering is proposed.Firstly,it clusters the traces with high behavioral similarity in the event log into several sub-logs according to the Levenshtein editing distance between traces.Then,it traverses each sub log,which selects the representative traces in the sub logs by using the methods of high frequency in the cluster and cluster center to form a set of candidate traces.Furthermore,it takes advantage of the optimal alignment technique to construct the behavior subset of the model,which calculates the degree of fit with the complete event log to get the approximate consistency and its upper and lower bounds.Finally,it carries out the simulation experiment of real event log,which verifies the superiority of this method from two aspects of accuracy and time efficiency.(3)Aiming at the problem that the existing prediction methods lack consideration of weak order relationship between log activities,a prediction method based on log behavior relationship and association analysis is proposed.Firstly,the log probability matrix is constructed based on the direct successor relationship of log activities,and a new algorithm is proposed to extract the profile relationship of log behavior from the matrix,which can effectively identify the complex structures such as loop,selection and parallelism;Then the behavior rule base is generated from the historical event log in combination with the log behavior relationship,the test data is converted into event flow,the prefix activities of the event flow are controlled by setting the backtracking window,and the association analysis and prediction are carried out by using the behavior relationship between prefix activities;Finally,the superiority of this method is verified by simulation experiments in terms of accuracy.Figure [25] Table [13] Reference [81]...