Norman Fenton research interests and vision
- Overall goal of my research has been to improve the way complex,
software-intensive systems are built and analysed. Chronologically this has focused on:
software metrics, software engineering methods and standards, and risk assessment.
- Much of my recent work has focused on risk assessment and decision
support under uncertainty, since I regard this as one of the most important
challenges in building critical systems. This built on my long-term research in software
measurement, but required us to meet new challenges, most notably to find methods for
combining and reasoning about diverse types of evidence (much of which is uncertain).
Hence my recent focus on Bayesian networks (BNs), which we identified as the most
suitable method for such reasoning, and multi-criteria decision aid.
- Our work in BNs has gained much international attention. We have
developed and applied BNs to predict the safety and reliability of software-intensive and
traditional mechanical systems, with some success. We have developed decision support
systems, sitting on top of BNs, which provide managers of complex systems with the
capability to identify and reduce project risks. In the process have built what we believe
are the largest BNs ever. Our projects have demonstrated that engineers and managers can
benefit from BNs without any need for specialist education in statistics and probability.
- In applying BNs to industrial-size problems we have had to solve
research problems not even considered by other more long-term BN researchers. For example,
we have developed methods and tools to help build large BN graph topologies and produce
large conditional probability tables from smaller sets of probability distributions.
- Although the main focus has been on critical systems, our research
has also included such diverse applications as reasoning about evidence in legal
arguments, and predicting football results.
- As well as applying BN technology to real problems our future
research work will continue to tackle issues arising from building decision support
systems for large-scale problems. This involves further research into large-scale BNs
(such as improving information visualisation and looking at ways of investigating the
sensitivity of results to changes in assumption), data-mining, and alternative methods of
uncertainty and decision support
- My early work in software metrics established a rigorous foundation
for the entire subject area (for example, using ideas from measurement theory). The
framework I proposed for software metrics has been widely adopted. However, increasingly,
it is clear that software metrics are primarily a tool for managing risk. Hence my current
emphasis in software metrics is also geared toward their use in software project risk
assessment and risk reduction. My very recent work in this area has developed improved
methods for predicting and managing software resources.
- I have a long-term interest in improving the empirical basis for
software engineering. Hence, much of my work has been concerned with both improving the
experimental basis, and also conducting experiments and major case studies to evaluate the
effectiveness of specific software engineering techniques. For example, I have conducted a
number of experiments to evaluate formal methods.
- Our best and most interesting research has arisen from trying to
solve real problems from commercial collaborators. Hence I have firm belief in the
overriding importance of problem-driven (rather than academic-driven) research. Much
software engineering research in the UK has focused on techniques that are only feasibly
applicable to small-scale systems. Hence, much of the research has not addressed the real
needs of system builders. Since computer science research is nothing if it cannot be
applied to realistic systems, I shall continue to focus only on research problems of
industrial relevance.
Return to Norman Fenton's homepage
Return to Norman Fenton's CV
Return to Norman Fenton's projects
page