Research On Automatic Modeling Oriented Scientific Workflow Recommendation Method

Scientific workflow is an automatic execution process of scientific computing,which has been widely utilized in the scientific computing fields of physics,astronomy,cheminformatics,bioinformatics and life sciences,but scientific workflow modeling is time-consuming and error-prone.In addition,the existing online scientific workflow repositories can support the publishing and sharing of scientific workflow,and these scientific workflows or their parts can meet some new modeling requirements.Therefore,based on the existing scientific workflow models in these repositories,scientific workflow recommendation research can be carried out to construct scientific workflow automatically.To improve the modeling efficiency of scientific workflow,this paper studies the recommendation method of scientific workflow for automatic modeling.The main contributions of this paper are as follows:(1)A scientific workflow recommendation method(LHWT)fused hierarchical structure and tag information is proposed.In this method,scientific workflows are transformed into the corresponding hierarchical structures,and the weighted mechanism is used to fuse the similarity of scientific workflow in editing distance and tags for scientific workflow recommendation.(2)A scientific workflow recommendation method(HDSWR)based on heterogeneous information network(HIN)and density peak clustering(DPC)is proposed.In this method,different data objects and underlying logical relationships in scientific workflow are organized into a HIN,and the similarity of scientific workflow in terms of meta path is used to calculate the similarity between scientific workflows effectively and improve the recommendation effect of scientific workflow model.(3)A scientific workflow activity recommendation method(SWAR)based on workflow segments is proposed.The concept of scientific workflow segment is introduced in this method,and graph mining technology g Span and graph decomposition algorithm are utilized to mine scientific workflow segments.Based on semantic similarity and distance measurement,a group of activities are recommended as the candidates of the next activity for the current under constructing scientific workflow to improve the efficiency of scientific workflow modeling.