The Business Process Prediction Based On Machine Learning

Process mining is a combination of business process and data mining.Today,the research of process mining focuses on the business processes prediction.The main tasks of business process monitoring are:the prediction of the next activity,the prediction of the future path(continuation)with a running case,the prediction of the remaining cycle time,the prediction of the deadline violations and predicting the fulfillment of a property upon completion.This paper mainly proposes the methods to solve the problems of the prediction of the next activity and timestamp,the prediction of the future path(continuation)of a running case and the remaining cycle time,the prediction of the results,then packages these methods into a prediction module.Finally,the prediction module is applied to the actual application,and a prototype system of business process monitoring and prediction is constructed.This paper proposes two prediction models for three prediction tasks.One model is used to predict the results of the process.A method using the Long Short-Term Memory model in deep learning is proposed for predicting the results of the process.The main purpose is to combine the prediction problem of process results with the natural language processing and propose a new solution.Another model is used to predict the time and event activity of business process.This model could solve two tasks which are predicting the next activity and timestamp,predicting the future path(continuation)of a running case and the remaining cycle time.This paper proposes a idea that use the Gated Recurrent Unit,the Bidirectional RNNs,the Attention technique and the Word2vec embedding method in natural language processing for business process prediction.The aim of the approach is to find the right technologies for business process prediction tasks with natural language processing.After studying the forecasting methods of three predicting tasks in depth,this paper constructs a prototype system of business process monitoring and forecasting by using the predicting methods.This prototype system mainly obtains the monitoring information of the running status of the business process being executed,and can predict the process results,activities and time.The system aims to enable users to monitor the running status of business process in real time and optimize the process in time by predicting its follow-up.