Empirical Evaluation of Metrics to Assess Software Product Line Feature Model Usability

A Software product line represents systems having conceptual similarity. All the systems in the product line have commonalities and variability. Feature models are often used to represent this intrinsic commonality and variability. They have a tree-like structure. A feature model which is low in quality will have negative effect over all the products belonging to the product line. Thus, early indicators in the form of metrics are required to assess quality of feature models. Assessments of quality attributes will help in avoiding the consequences of inferior quality and faulty design at the later stages of production. Quality attributes are of two types: first type of attributes is internal which can be measured through product related features like length, complexity, efficiency etc. Second are external attributes which can be assessed once the product is fully functional like usability, reliability, maintainability etc. Usability is an external attribute which focuses on easy and efficient product usage. In reference to feature models usability is an essential quality to be possessed. In this paper, we intend to validate an existing structural metrics for software product line feature models. In our research, we try to examine whether the available metrics are fair indicators of the three main sub characteristics of usability viz. learnability, understandability and communicativeness. We try to analysis whether these existing metrics for feature models have a correlation with feature model usability. For this we have employed statistical correlation techniques. Results obtained from the empirical validation shows that the metrics are correlated to the subjective perception regarding the usability of the feature models. Keywords-Software product lines, feature models, structural complexity metrics, empirical validation.