Assessing Software Product Maintainability Based on Class-Level Structural Measures

Abstract

A number of structural measures have been suggested to support the assessment and prediction of software quality attributes. The aim of our study is to investigate how class-level measures of structural properties can be used to assess the maintainability of a software product as a whole. We survey, structure and discuss current practices on this topic, and apply alternative strategies on four functionally equivalent systems that were constructed as part of a multi-case study. In the absence of historical data needed to build statistically based prediction models, we apply elements of judgment in the assessment. We show how triangulation of alternative strategies as well as sensitivity analysis may increase the confidence in assessments that contain elements of judgment. This paper contributes to more systematic practices in the application of structural measures. Further research is needed to evaluate and improve the accuracy and precision of judgment-based strategies.