Fuzzy Set and Cache-based Approach for Bug Triaging
Ahmed Tamrawi, Tung Thanh Nguyen, Jafar Al-Kofahi, Tien N. Nguyen

Selection of Developers and Terms

To evaluate the impacts of both types of selection, we conducted another experiment and tuned the model with different sizes of developer cache and term cache to get the better results. For each subject system in Table 1, we ran Bugzie on all datasets with all combinations of the best values we discovered in the previous experiments as the model’s parameters/configurations.

Tables 4 and 5 show the accuracy and total processing time with different parameters for Eclipse and Firefox. As seen, Bugzie could be tuned to achieve very high levels of accuracy and efficiency.





Table 6 shows top-1 and top-5 best accuracy when we ran Bugzie with 4 types of configurations: base model with all developers and all terms (Column Base), the one with candidate selection (Column C.S.), the one with term selection (Column T.S.), and the one with both (Column Both). Generally, the top-5 accuracy achieves the best results in the range of 75-83% for all projects (except for NetBeans - 61.3%). The best results for top-1 accuracy are from 31-51%.