Workflow Exception Forecasting Method Based on SVM Theory

Based on the analysis of workflow exceptions and Petri net theory, the study presents an extended modeling method, Label Petri net, which is used to describe workflow more accurately with state information. With these state information and data, a new method applying Support Vector Machine (SVM) theory is put forward to forecast workflow exceptions. This method can forecast overall and local exceptions and attempt to provide evidence for handling exceptions and developing flexible workflow system. Finally, a supply chain workflow system case about a packaging factory shows the application of Label Petri net modeling method and describes computing process of SVM theory to forecast exception. The calculation results show different effect with various kernel functions, which is explained in this thesis. And the study gives some advices to choose kernels of SVM algorithm. The case demonstrates feasibility and effectiveness of this method in forecasting workflow exception.