Research On Graph-based Workflow Recommendation Using CNN

Business process management is the core of enterprise information system,and efficient and accurate business process modeling is the inevitable requirement of modern enterprises in dealing with the frequently changing market demand.As artificial modeling is complex and needs to invest a lot of manpower and resources,so the industry demand for auxiliary process modeling technology is becoming increasingly urgent.Business process recommendation is one of the most effective auxiliary modeling techniques at home and abroad.Traditional business process recommendation algorithms are mainly based on the idea of similarity matching.The GED_based recommendation algorihm takes the graph edit distance of business processes as the similarity metric,while SED_based recommendation algorithm firstly encodes the business processes with minimum DFS code and then takes the string edit distance as the similarity metric.However,these algorithms generally face the challenges of poor real-time and low accuracy.In this paper,we propose a business process recommendation technique based on convolutional neural network.In this paper,we firstly give a standard and clear definition of business process recommendation,based on which the overall architecture of the existing business process recommendation system is studied,and its off-line mining and on-line recommendation module are analyzed emphatically.Then,we analyze a real business process data set,and summarize the data preprocessing method and data characteristics.Finally,based on the shortcomings of traditional algorithms and the characteristics of data sets,this paper designs a business process recommendation algorithm based on convolutional neural network and applies it to the preprocessed real data set.The results of cross validation experiments show that the algorithm proposed in this paper has achieved good results in both time consume and accuracy.