Experimental Study of Quantitative Analysis of Maintenance Effort Using Program Slicing-Based Metrics

During the software development lifecycle, studies have shown that over 75% of project costs originate from the maintenance phase. Analysis of the processes within the maintenance phase could prove beneficial since most maintenance activities revolve around source code. Accurate estimations of the maintenance effort spent on code changes would enable cost effective management of resources. In this research, we investigate a quantitative approach to express maintenance effort, for which a set of program-sliced metrics is proposed. Using the time to resolve an issue as a measure of maintenance effort, we evaluated our proposed metrics against the basic code-based metrics Lines of Code and McCabe´s Cyclomatic Complexity. To eliminate outside factors, we performed an experimental case study on a set of pre-defined maintenance activities. Results suggest that program slicing metrics have the strongest correlation with maintenance effort, exhibiting a moderate degree of correlation with maintenance effort. In contrast, Lines of Code has a weak correlation with maintenance effort. This study contributes to our ongoing research into the analysis of maintenance processes.