Next: Monitoring the testing process
Up: Future Work
Previous: Future Work
As discussed in chapter , the case study at Beckman
was confined to a single group of developers, a single Bayesian
model of confidence-level uncertainty, and a single scenario of
change with associated questions. We are very interested in
expanding the scope of our field studies, both within Beckman
as well as with other software development organizations.
A more comprehensive study than the one reported in this
dissertation would include one or more of the
following enhancements:
-
Measure the effects of using the Bayesian models over time,
especially its impacts on the quality of human judgment and
decision making.
-
Construct additional Bayesian models for additional artifact
networks or to model other situations or properties.
We are particularly interested in constructing Bayesian models
at different levels of granularity, comparing the efficacy of
resulting models, and, eventually, combining Bayesian models
with different granularities.
-
Afford automated support for definition and revision of
software belief networks. Of special interest are UI design
issues in constructing belief-network software as well as
integration
of Bayesian capabilities with existing tools, e.g., REBUS,
DOORS, and RequisitePro.
-
Finally, this dissertation focused on incorporation of
subjective information into software models. We are
interested in combining subjective and objective data
in future models. Specifically, we wish to include
project data such as code quality and developer
productivity, failure rate and severity, etc., in our
models as factors that influence confidence and,
ultimately, human judgment and decision making.
Hadar Ziv
Fri Jun 20 16:25:19 PDT 1997