Weighted Multi-Factor Multi-Layer Identification of Potential Causes for Events of Interest in Software Repositories

Change labelling is a fundamental challenge in software evolution. Certain kinds of changes can be labeled based on directly measurable characteristics. Labels for other kinds of changes, such as changes causing subsequent fixes, need to be estimated retrospectively. In this article we present a weight-based approach for identifying potential causes for events of interest based on a cause-fix graph supporting multiple factors, such as causing a fix or a refactoring, and multiple layers reflecting di↵erent levels of granularity, such as project, file, class, method. We outline di↵erent strategies that can be employed to refine the weights distribution across the di↵erent layers in order to obtain more specific labelling at finer levels of granularity.