Binary classifiers are used for bug deduplication in software development. However, the precision and recall metrics of individual classifiers may be insufficient for particular use cases. Selecting an appropriate binary classifier can improve bug deduplication, e.g., high precision can ensure that duplicate bugs are identified (and thus not analyzed), while providing sufficient recall (to eliminate redundant analysis of bugs that are duplicates). This disclosure describes automated techniques to identify the most appropriate classification model from a set of models. Two different functions - F-beta score and weighted sum score - are evaluated for different performance thresholds. The model and threshold combination with the highest score is selected and used for bug deduplication.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.