Next event prediction based on multiview Transformer
Predictive business process monitoring relied on historical data stored in event logs to predict future trends in currently executing processes.Existing methods based on deep learning often only considered the activity and timestamp of the event,ignoring other attributes of the event,resulting in inaccurate prediction of the next event.To address this problem,a multi-view Transformer model was proposed for prediction,aiming to improve the prediction accuracy of the next event.This model was based on various event information recorded in event logs and incorporates a self-attention mechanism to establish complex dependencies between event sequences and corresponding outputs,making more comprehensive use of information in the event log.The experimental results on 4 real-life event datasets showed that the proposed method improved the accuracy of the task of predicting the next event to a certain extent and more accurately predicts events in future business processes.
business processdeep learningpredictive business process monitoringself-attention mechanismmulti-viewTransformer
万方数据
维普
