IMPRESS (Improving the Software Process Using Bayesian Nets)
EPSRC Project GR/L06683. 1 Jan 1997 -
Note: This Project has been transfered to Queen Mary and Westfield College, University of London, since 1st April 2000. Contact Dr Martin Neil (martin@dcs.qmw.ac.uk) for further information.
Funding to CSR, City University: £220,792
Summary: Integrating novel research in safety and dependability assessment into mainstream software engineering process improvement. Developing a tool to predict specific software quality measures directly related to reliability and maintainability.
Objectives:
- To use our past research results in dependability argumentation using Bayesian Belief Network's (BBN's) and decision analysis techniques to control and measure software quality,
- to develop a BBN approach to statistical control of software processes and for prediction of the industry standard quality measure, defect density,
- to implement these new BBN's in a demonstrator prototype for promotion within the research community and dissemination to industry.
Approach:
- A concise IMPRESS framework is being developed with the aim of recording the steps needed in quality decision making and quality argumentation.
- Incorporating variables drawn from quality modelling, defect density prediction and process improvement, two BBN templates are to be developed predicting defect density and maintainability.
- Specifying alternative methods for software Statistical Process Control
- Developing a prototype demonstrator tool from an existing BBN tool.
Results: Currently in mainstream software engineering, including the commercial sector, there is a continuing reliance on ill-conceived, ad-hoc methods for assessing the quality of processes and products. IMPRESS will provide a prototype tool that incorporates models of software quality, defect analysis and statistical process control for prediction of software quality measures directly related to reliability and maintainability.
Impact: In the long term, any industry which relies on high quality software products may benefit from this research. The medium-term beneficiaries will be software quality managers and independent assessors of software processes and products. Also consultants, tool developers and vendors. The short-term beneficiaries will be researchers in software quality and process improvement.
European Partners: HUGIN Expert A/S (Denmark).
CSR Personnel: Professor Fenton, Professor Littlewood, Dr. Neil, Professor Strigini, Mr. Lewis, and Mr. Makwana.