Multi-angle business process next activity prediction method based on CNN-BiLSTM hybrid model

• Main Authors: SUN XIAOXIAO, YU DONGJIN, YING YUKE
• Format: Patent
• Language: Chinese; English
• Published: 08.12.2020
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; DATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FORADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORYOR FORECASTING PURPOSES; ELECTRIC DIGITAL DATA PROCESSING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE,COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTINGPURPOSES, NOT OTHERWISE PROVIDED FOR

The invention discloses a multi-angle business process next activity prediction method based on a CNN-BiLSTM hybrid model. According to the invention, the method comprises the steps: extracting threenext candidate activity attributes on the basis of relationships among event activities, attribute similarity and positions of events in instances, and taking the extracted attributes and basic attributes and time attributes of a data set as inputs of a hybrid model of CNNBiLSTM to carry out prediction of the next activity. The method has the characteristics of high prediction precision and wide applicability, and can effectively solve the prediction problem of the next activity in some complex scenes, thereby providing effective information for a process executor to prevent the occurrence ofconditions such as abnormal process execution sequence and the like. 本发明公开了一种基于CNN-BiLSTM混合模型的多角度业务流程下一活动预测方法。该方法基于事件活动间的关系、属性相似性、事件在实例内的位置提取了三个下一候选活动属性，并将提取出的属性和数据集的基本属性、时间属性一同作为CNN-BiLSTM的混合模型的输入来开展下一活动的预测。这种方法具有预测精度高、适用性广泛的