Research On Data-Aware Business Process Prediction Approach

Business process management(BPM)is a discipline aimed at improving the efficiency of business processes and involved the design,deployment,execution and other phases of processes.With the scale and complexity of business processes in organizations,the role of BPM becomes more and more important,and business process prediction is an important branch of BPM.Business process prediction can monitor and predict the events that are being executed in real time.The predict targets include process results,completion time,and subsequent activity sequences,etc.,so that this technology can help process participants to take measures in advance to prevent or reduce losses in the future.In recent years,the transformation of centralized business processes to distributed deployment has become an inevitable trend.In a distributed environment,event logs are stored in a distributed manner,which results in a large cost and poor timeliness to obtain a complete event sequence.Thus,event sequence-based business process prediction encountered a bottleneck.In order to get rid of the limitation of the process prediction task due to the difficulty of obtaining event sequences,this paper proposes the data-aware business process prediction method(DABPP).DABPP is an event-based business process result prediction method that departs from the concept of cases in traditional business processes and is more suitable for distributed deployment environment.The main work and innovation of this paper are as follows:This paper conceptually divides events and their additional data,and proposes the definition of the result stage of the process.These concepts are the theoretical basis of the process prediction method based on events.Event and event-related additional data were used to predict the results of the process in stages using probabilistic automata combined with neural networks.This paper conducts further analysis of the activity attributes of the prediction results,and analyzes the degree of influence of each attribute on the process results in each activity,so as to give suggestions for the optimization of business process execution.Finally,this paper uses the real event log BPI2017 and the comparative baseline to evaluate the predictive performance of DABPP method and give a case study of the activity attributes analysis method.