Publication: A bayesian model for change impact on SW estimations
Loading...
Identifiers
Publication date
2015-10
Defense date
2015-12-18
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
(….) This thesis focuses on these problems: first, the need to give accurate estimations to drive the decision process; second, the need to identify and perform a mental mapping of the elements involved in the design decisions; and third, the need to maintain this “network” of decisions in such a way that it is being shared by the stakeholders within a project.
In order to tackle these problems, we will use Bayesian Belief Networks (in short, BBN) for representing the main concepts related to a given project, their causal relationships, as well as the associated conditional probabilities. BBNs are well-defined analysis techniques based on probability calculus that have been used for estimations in multiple areas (Kjærulff, et al., 2005). The main advantage of using BBN for project estimations and measuring change impact is that they allow the estimations to be based on uncertainty and incompleteness of the input parameters. In addition, BBN techniques allow software engineers to use an explicit representation of the causal relationships between the relevant project attributes.
BBN estimations also allow us further refinement once these parameters are known.
The estimation and change impact problems are related to three different areas: change impact analysis techniques allow to determine the software artifacts involved in change before the change is made; software estimations and, in particular, the use of Bayesian Belief Networks for software estimations provide the basis for estimating in advance the effort, risk, and/or quality associated; and design rationale gives us the required background for analyzing the mental process associated to software changes. In the following chapters we will provide an overview of the state of the art in these techniques.
Description
Keywords
Bayesian believe networks, Software estimations