Objective
The MODIST aim is quite novel: To achieve measurable gains in the resultant quality of distributed software development through new decision and risk management techniques. These techniques will make the causes of, and uncertainties associated with, variations in software quality visible to managers and engineers. To explain these variations, MODIST is developing a Bayesian Process Control (BPC) approach for software (using Bayesian Belief Network models). The BPC based approach comprises of two method components: MODIST development (for creating BBN software quality models); and MODIST application (for using the models) for quality assessment within different development scenarios. Also, a decision support demonstrator tool will integrate the methods and models to identify and improve software quality in distributed development situations.
OBJECTIVES
MODIST's aim is to achieve measurable gains in the quality of distributed software development through decision and risk management techniques.
The objectives are:
- To develop a Bayesian process control (BPC) method (using Bayesian Belief Models)(MODIST development method) for creating software quality models that incorporate the uncertainty and risks involved to support different development scenarios;
- To develop a decision support process (MODIST application method) for using the BPC based software quality models;
- To capture the MODIST application method tailorable to different distributed software development scenarios in the form of a MODIST demonstrator tool;
- To define and execute an effective MODIST validation methodology involving detailed trials and industry wide expert assessment.
DESCRIPTION OF WORK
The project will develop models, methods and a demonstrator tool for BPC, a statistical process control (SPC) analogy, for distributed software development based on Bayesian Belief Networks (BBNs).The project includes 5 leading system and software based organisations including 3 users and 2 technologists. The users are in vastly different markets and involved in global software development. The technologists cover specialisms in software engineering and quality, process management, risk management and in causal modelling with BBNs. MODIST will create methods for: model development (build models to explain quality variation and uncertainty); and model application (use models to assess quality for development or software process improvement (SPI)). Model development methods will involve capturing domain and engineering understanding. Model application methods involve using models to understand about software decision making to be tailored to distributed development scenarios. The methods and tool will provide software managers and engineers with an innovative and easy to use decision support system that may be integrated within realistic software project decision making processes. The methods/demonstrator tool are innovative not just in their application domain, but also in the scale of BBN models that are to be built, and in the application of new, usually impossible, SPC/BPC methods of software quality assessment. The project will consist of 6 technical workpackages over 30 months: software scenarios, method requirements, domain measurement, methods development, validation (via trials and user group) and finally consolidation. The workplan strategy is based on: initial domain understanding; early prototyping; extensive validation; and final consolidation of the method and tool based on the available validation evidence.
Fields of science
Topic(s)
Data not availableCall for proposal
Data not availableFunding Scheme
CSC - Cost-sharing contractsCoordinator
LONDON
United Kingdom