Research On Context-oriented Scientinc Workflow Modeling And Its Quantitive Analysis

The lack of flexibility is an inhibitor in the use of scientific workflow (SWF) systems in order to transform hypotheses into scientific discoveries. Despite the fact that an increasingly broad view is taken into account to address the issue of SWF flexibility, there is a notable absence of methodology for modeling SWF in a uniform representation and for encapsulating it as integrated SWF constructs to support scientists'decision making process. Moreover, existing SWF systems rely much on available process and data, which lacks a user-centered approach to capture the user's particular demands in a given situation.The main objective of the dissertation is to investigate a uniform formalism for SWF modeling, which could represent scientists'knowledge, reasoning and contextual information in a global viewpoint. Firstly, we proposed that a real use of SWF requires the introduction of the concept of "context". Secondly, we proposed a Context-oriented Framework for Scientific WorkFlow (CoF4SWF) to make contexts explicit in SWF, provide decision support for scientists, as well as enable the incremental acquisition process of novel knowledge, by modeling tools, such as context-oriented SWF metamodel and Contextual Graphs (CxGs). Thirdly, we built the "decision support model for making context explicit" by AHP model, involving three methods:(1) the user acquires decision support from the system;(2) the system generates all possible SWF models automatically;(3) the user and the system collaborately design SWF models through the interactivity. Furthermore, we implemented CoF4SWF methodology by a functioning prototype called Context-oriented SWF (CxSWF). Moreover, CxSWF prototype was evaluated and testified through a real world application "virtual screening research on avian influenza H5N1virus" as an implementation framework. Finally, the results from both the quantitative analysis of questionnaires and twofold evaluation processes illustrated that initial hypotheses of the methodology and the application have been reached in principle.The main contributions of the thesis are:Firstly, a context-oriented methodology for studying SWF modeling are presented, which has also been evaluated through the quantitative analysis of the application. Secondly, three approaches of making contexts explicit in SWF decision support model are identified. Thirdly, the role of contextual graphs in SWF modeling is investigated. Furthermore, the originality of the thesis stands on:(1) representing and modeling SWF flexibility with its contexts to enable adaptive decision making;(2) making contexts as the status of knowledge to be built up for SWF repository by the processes of both practice learning and incremental knowledge acquisition; and (3) sharing of SWF models with their contexts as one of complementary outputs for scientific discoveries.