Learning to Aggregate: An Automated Aggregation Method for Software Quality Model

Quality models are regarded as a well-acceptedapproach for assessing high-level abstract quality characteristics(e.g., maintainability) by aggregation from low-level metrics. However, most of the existing quality models adopt the weightedlinear aggregation method which suffers from a lack of consensusin how to decide the correct weights. To address this issue, wepresent an automated aggregation method which adopts a kind ofprobabilistic weight instead of the subjective weight in previousaggregation methods. In particular, we utilize a topic modelingtechnique to estimate the probabilistic weight by learning froma software benchmark. In this manner, our approach can enableautomated quality assessment by using the learned knowledgewithout manual effort. In addition, we conduct an application onthe maintainability assessment of the systems in our benchmark. The result shows that our approach can reveal the maintainabilitywell through a correlation analysis with the changed lines of code.