What is DEUM?
DEUM analyzes and evaluates the feasibility of unlabeled data for unsupervised defect prediction models (USDP).
We identified potential indicators for unlabeled data evaluation and implemented the DEUM approach.
CLA/CLAMI is one of unsupervised defect prediction models that clusters the defect data and labels the data.
For now, DEUM is available for CLA/CLAMI models using a data indicator.
DEUM will be extended for other USDP models, later. (You can access Javadoc here: http://isel.lifove.net/Javadoc/clamiJavadoc/index.html)
The indicator used in DEUM is "GIR" which stands for Group Instance Rate.
Group stands for number of groups that is computed by the number of group of metrics that has high correlation with other metrics.
Do you want to check if data is feasible for CLA/CLAMI ?