Business Process Prediction Based On Encoding And Decoding

The execution of business process produces a large number of event logs,which provides an important basis for business process mining.Process prediction is a hot issue in business process mining.The goal is to analyze the current process instances and infer their possible execution status in the future.Business process prediction includes activity prediction,time prediction and violation prediction.Prediction methods are mainly divided into model-based prediction and deep learning prediction.In recent years,the prediction based on deep learning is mostly improved on the basis of the LSTM neural network model.There are still deficiencies in the data coding mode and activity sequence training,which do not take into account the semantic information of the data itself and do not make full use of the correlation between the activity sequences.In this paper,glove model is used to encode event log data,and attention mechanism is used to build a business process prediction model from coding to decoding to solve the problem of activity and time prediction in business process.The main work is as follows.First,considering the semantic relationship between the activity sequence data,we use the glove model to encode the activity sequence in the event log.Based on the structure characteristics of encoding and decoding in neural network,with the attention mechanism,a business process prediction model EDPM from encoding to decoding is established.Log vectors with different tag values are generated for the next activity prediction and the subsequent activity prediction,which are input into the prediction model for training to get the final activity prediction model.Secondly,considering the continuous value of time series,the initial execution time of activity series in each track is discretized.Combined with the activity sequence,it is added to the event log as the tag value of its input eigenvector.The constructed event log is encoded to get the log vector and input into the EDPM prediction model for training to get the final time prediction model.Finally,the model proposed in this paper is tested on two public datasets.By comparing the experimental results of different scale datasets,the validity of the proposed model is verified.Compared with the existing prediction model,the model proposed in this paper can improve the accuracy of prediction.