Research On Remaining Time Prediction Algorithm Based On Process Mining

Process mining technology can extract valuable information from event logs generated by various modern information systems.This technology provides a new way to improve processed in various application fields.Predictive monitoring of business process is a new research focus in the field of business process management under the development trend of global intelligent manufacturing,involving theories and technologies in the fields of management,operations research and artificial intelligence.The goal of predictive monitoring of business process is to perform predictive analysis of running process instances,such as giving the next activity to be performed,the best person to perform an activity,or a running business within a specified period of time whether it is completed on time,the time required,etc.Compared with traditional process monitoring methods based on dashboards and reports,the advantage of predictive monitoring of business process is not only that it can monitor the execution and running status of each process instances in real time,but also that if can intelligently perform future behaviors.In actual social production life,we need practical and reliable prediction information in many cases,for example,when an insurance customer calls the insurance company to ask for the claim payment information,the insurance company will give a relatively accurate answer.To this end,this paper proposes a multiple transition system model for the online process remaining time prediction problem,and then proposes an abstract selection method based on data purity,which automatically selects the abstract with the highest data purity for remaining time prediction.After in-depth research on predictive monitoring of business process,this paper uses the realized forecasting method to design and develop a corresponding process mining and time forecasting system.This system mainly predicts the remaining time of the running business process,and aims to allow users to conveniently and effectively monitor the running status of the business process and adjust them in a timely manner with the forecast results.The method in this paper breaks the abstract limitations of the traditional transition system method and has a certain theoretically innovative.Experiments show that for real event logs,the method proposed in this paper is superior to the benchmark method in terms of stability and accuracy.After considering the purity of the data,the accuracy of the remaining time prediction has been improved to a certain extent.