A <i>Bayesian network</i> approach to assist on the interpretation of software metrics

Abstract

Despite the quantity of software metrics that has been proposed, their adoption and application by practitioners has been limited. A challenge to their use is to interpret them to perform assessments and predictions. The existing approaches to assist with their interpretation consists of defining thresholds to determine whether the value of a metric is acceptable. These approaches are not enough to ensure a correct metrics' interpretation, because they ignore risks and other subjective factors that influence the measurement process. This might affect the metrics' interpretation, and consequently, the manager's decision. To minimize wrong decisions based on software metrics, we present a method to construct Bayesian networks to assist on metric interpretation considering these risks. We successfully validated the method with a case study performed in three software development projects. We concluded that it is a promising approach to assist practitioners to interpret metrics and support software projects managerial decision-making.