A Model-Driven Engineering Framework for Computational Replicability and Reproducibility
Type of DegreeDissertation
MetadataShow full item record
The scientific community currently suffers from a lack of verification of results obtained from many well-intentioned researchers. Part of the reason for this lack is the difficulty in reproducing experiments performed by these researchers. One of the leading causes of this problem is replicating the computer models used in these experiments. Also, some reviewers may not be familiar with the tools, modeling environments, and languages used by the original researchers. Currently, the solutions to this problem typically rely on the ability and willingness of the original researchers to completely capture all pertinent details of the experiment and the reviewers to have a great deal of familiarity with both the nature of the experiment and the tools used for experimentation including the source modeling environment. Unfortunately, if one or more of these aspects is lacking, then a reliable review of the original experiment may be extremely difficult if not impossible. As such, a need has arisen for the ability to reliably transform computer models from one modeling environment to another and the maintenance of platform-neutral representations for use in future experiments. A process for transforming a model that is specific to a particular platform into a representation that is not dependent on a specific platform is necessary. Creating such a definition of the model can then be transformed into other environments for validation of the results obtained from simulations using the original model. We present a solution using a hybrid of two transformation technologies to facilitate the execution of such a process. We show how the process has been used to successfully produce a Platform Independent Model from the essential components of a model developed in a Platform specific environment. We present the details of how we confirmed that this Platform Independent representation can be used to generate a second model in the original platform and the results from executing this second model are the same as the original model; thereby showing that the models represent the same model structurally. Additionally, we present the work done to develop a web-based user interface to perform the necessary model transformations. This process will be a useful tool to assist with the ability of researchers to develop and execute reproducible simulation experiments. By using this tool to confirm the results obtained from simulation experiments, the results of these experiments will be more reliable, generating increased credibility for the scientific community.