Business Process Prediction Based On Event Log Sampling And Prefix Tree Enhancement

Predictive business process monitoring mainly uses the data of the process execution that has already occurred,and predicts how the ongoing process will be performed in the future by establishing a deep learning model.However,the most advanced predictive monitoring methods need to train complex machine learning models.In big data In the background,more and more data increases the training time of the model,which is often inefficient based on hardware and time constraints.The focus of this study was on how to maintain or even improve prediction accuracy without reducing it as much as possible,on the basis of reducing training time.Based on the original prediction model,this paper uses the event log sampling method to improve the training speed,and uses the prefix tree to represent the behavioral relationship between activities for prediction enhancement to improve the prediction accuracy of the model.The main research content of this paper is as follows.(1)Aiming at the problem that existing hardware and time constraints hinder the application of machine learning-based technology in real business processes,a predictive process monitoring method based on event log sampling is proposed.By sampling the event log,two sampling methods are adopted in this paper,which are equal-proportion sampling and sampling according to the importance of traces.And use the sampled event log to train the prediction model.The experimental results show that our sampling method can greatly improve the training speed while maintaining sufficient prediction accuracy.(2)In order to use the hidden behavior relationship between logs for predictive process monitoring,a business process prediction enhancement technology based on prefix tree is proposed.Helps improve the quality of predictive models by mining behavioral relationships between logs during process execution.The behavior relationship between the excavated logs is represented by a prefix tree,and the existing deep learning-based business process prediction model is used to provide decision support through the prefix tree structure to filter the prediction results that conform to the behavior relationship in the result prediction stage.Improve the accuracy of the prediction results and compare them with the baseline method in the event log.The experimental results have improved the prediction accuracy in predicting the next activity and predicting the suffix.(3)Aiming at the problem that event log sampling will reduce the prediction accuracy of the model,a combination algorithm based on event log sampling and prefix tree prediction enhancement is proposed.The problem of model prediction accuracy degradation is solved by prefix tree enhancement method.Before training the model,sample the four event logs and use the sampled logs for training,and use the prefix tree enhancement technology to improve the prediction accuracy of the trained model in the prediction stage.The experimental results show that this method can greatly improve the training speed.In some cases,the prediction accuracy is not reduced and most of the prediction results are improved.Figure [19] Table [16] Reference [91]...