Reviewer 3

[WeiFoo]

Comments on this reviewer : you're a good researcher!

• [x] Q2: lessmann did use some search algorithms to find optimal parameters, should clarify and should discuss insufficiency of such settings.
• [ ] figure 5, some precision results listed in Figure 5 are very low (< 20%), even after tuning. The authors may discuss more about that.
  • tuning can't guarantee to find a better solution that precision > 20%, probably because of the nature of the data sets
• [ ] A further interesting question that the authors may consider (in future) is to investigate whether the defect prediction using DE belongs to convex optimization (if so, then using metaheuristics makes perfect sense).
  • not considered
• [x] The authors claimed that the default values are "standard". But it is hard to find literature using these "standard default values" for defect prediction. need more explainations
• [x] the authors may explain why they choose the defined tuning ranges for the parameters. Are they adjustable for the evaluated case studies?

[timm]

there are papers showing that grid search is NOT better than other optimizers. can we find them?

also, not sure what this reviewer said we were doing wrong... i think they were actually our advocates

btw.. getting 3 reviews meant we made it into round2 of the discussion. so we were close on this one

[WeiFoo]

actually, this reviewer did support our work and most of his feedbacks were suggestions. he bought our idea.

[WeiFoo]

I will close this one, I'm done