Source Code Metrics and Maintainability: A Case Study

Towards Building Method Level Maintainability Models Based on Expert Evaluations

“…One method was evaluated by only one participant and each participant evaluated 10 methods. The results pointed out that there was a large deviation in the judgments of the sub-characteristics which a ected the e ciency of the built prediction models [11]. The cause of the large deviation can be that di erent experts might have di erent subjective scales and di erent interpretation of the same quality concepts.…”

“…As a sequel of our previous work [11] 268 participants took part in the experiment where the three surveys introduced in Section 1 were performed and more than 150 000 questions were answered (for evaluation statistics see Table 1). The participants had to score the sub-characteristics of maintainability de ned by the ISO/IEC 9126 standard (analyzability, changeability, testability, stability) and a new quality attribute, comprehensibility, introduced by us [11].…”

“…The participants had to score the sub-characteristics of maintainability de ned by the ISO/IEC 9126 standard (analyzability, changeability, testability, stability) and a new quality attribute, comprehensibility, introduced by us [11]. Besides these quality attributes, the students had to evaluate the maintainability of the methods as well.…”

“…The evaluation has been performed with the help of our online survey system called Metric Evaluation Framework. For details about the application and technical questions refer to our earlier paper [11].…”

“…In our previous work [11] we built a maintainability model based on the ISO/IEC 9126 [13] standard by applying classi cation algorithms (using source code metrics as predictors) on manually labeled methods. The labeling of 350 Java methods was performed by 35 IT experts in such a way that each expert evaluated 10 di erent methods.…”