Granular Modelling in Software Engineering
Models of Software Engineering aimed both at software products and processes
are crucial in assuring software quality. Being inherently human-centric,
granular modelling can potentially help to address several critical methodological
and applied issues
- Identification of important tradeoffs between numeric (e.g., times of failure
and quantitative assessments (like software cost predictions coming from experts)
in building granular models satisfying fundamental requirements of their accuracy,
relevance, and transparency. The issues of this nature arise in the framework of
structure-free models such as association analysis, relational structures and genetically
optimised testing models that exploit various aspects of information granularity in
forming test sets;
- Combining information granules being expressed in different formal frameworks of
granular computing (e.g., fuzzy sets, rough sets and probabilistic granules). E.g.,
this concerns a broad class of models of software reliability where heterogeneous
information become available within the context of probabilistic models and need to
be combined with expert-driven models of software risk assessment;
- Building hierarchies of software models and constructing communication mechanisms
that help refine structures at various level of granularity.