For longevity, a software company should adopt value-based decision making, in which the focus is on overall value creation rather than the required efforts and costs alone. The paper introduces concepts behind value-based decision making, and illustrates them through a case study on feature selection “for the next sprint of an Internet of Things (IoT) project.”
The authors identify why a Bayesian network is a suitable technique for value estimation in the context of software engineering, particularly for effort estimation, quality prediction, and requirements engineering enhancements: “it has been successfully employed ... in several complex domains (e.g. genetics, speech recognition, medical diagnosis, software project management)”; “it supports reasoning under uncertainty”; it is suitable for “the representation of well-characterized uncertainty and ... decision options”; “it enables reasoning under uncertainty”; and it has “a sound [theoretical] basis in Bayesian probability.” In the context of their research, Bayesian network models are used to represent domain knowledge in terms of factors thought to be important.
This well-written paper describes a framework consisting of Bayesian network models, a knowledge base, and a decision algorithm (VALUE). The embedded process consists of: eliciting company-specific value factors; employing artificial intelligence (AI) techniques and tools in decision-making meetings; the semi-automatic generation of a probabilistic estimation model; validating the value estimation model; and an add-on value estimation model for decision making.
This is a quite useful paper for researchers and developers in the software engineering field.