Scientific Workflow Recommendation Based On Multiple Sub-requirements

The emergence of virtual research environment and virtual platform has had a great impact on the research environment and research process of researchers.In order to guarantee and promote scientific research collaboration in virtual knowledge environment,the scientific workflow has emerged.Scientific workflow is a kind of technology that helps researchers manage the complex process of scientific computation.It mainly consists of activities and the links between activities.In general,building a new scientific workflow from scratch is a knowledge-intensive,timeconsuming,and error-prone task.Therefore,it is very important to study the reusability of scientific workflow.At present,there have been a lot of researches on the reusability of scientific workflow.It is worth noting that these researches require users to provide clear and complete requirements description when recommending corresponding scientific workflow to users.However,in real life,it is easier and more likely for a user to provide multiple small sub-requirements and relationships between sub-requirements than a single complete,unambiguous requirement.Based on this,this thesis proposes a scientific workflow recommendation technology based on multiple sub-requirements to promote the reuse of scientific workflow.Specifically,1)For the diversified requirements of users,an activity network model representing the invocation relationship between activities is constructed to meet the diversified requirements of users.In addition,the activity network model can lay the foundation for sub-requirements matching and connecting the matching results of multiple sub-requirements.2)Based on the constructed activity network model,the precise matching technology is firstly used to match the multiple sub-requirements provided by user,and the inexact matching technology is used to match the sub-requirements that cannot be precisely matched,so as to find the scientific workflow fragment set satisfying the multiple sub-requirements of user.3)Based on the matching results of multiple sub-requirements,the constructed activity network model is reconstructed to find the shortest path between the matching results connecting multiple sub-requirements.Based on the reconstructed activity network model,a group Steiner tree algorithm is adopted for connecting the matching results of multiple sub-requirements.Multiple candidate scientific workflows corresponding to user requirements are generated,and then the generated candidate scientific workflows are sorted and recommended to user s.Extensive experiments are conducted on the my Experiment scientific workflow dataset in this thesis.The results show that the proposed technique is effective and accurate in most cases.