andre will reframe by end of end of May 23

• what is the average size of test suite in this domain?

• smaller stuides that have offered conclusion

• the soundness of DDR is not clear. The decision on whether a simple HPO can be used is based on the fact that at least 30% of the attributes are redundant. However, the rule does not consider the absolute number of attributes. Reducing the attributes from 100 to 60, is not the same as reducing from 1000 to 600. In the latter case, the problem is still complex.

• data covers this, not explained very well

• the paper claims to find results for SE problems, but it only considers a very specific type of problems, i.e., regression and classification problems. There are several SE problems that do not fit these categories: test case generation, test case minimization, refactoring, requirements prioritization, verification, etc. So, the paper should be restructured (title, abstract, introduction) to make it clear that the approach is not for all the SE problems, but for some specific ones. It is totally fine to consider a family of problems, but this should be clarified;

• jsut agree and reframe

• RQ1 is not properly articulated and answered. The selected datasets for SE and non-SE problems cannot be taken as representative of these two worlds. The paper should just state the RQ in terms of the type of problems it analyzes;

  • just focus on classification, regression, optimization
  • dont make a result.s jsut call it " somethign todocument and offer as a research chalelgne"

• 🅰 RQ1 is not properly articulated and answered. The selected datasets for SE and non-SE problems cannot be taken as representative of these two worlds. The paper should just state the RQ in terms of the type of problems it analyzes;

• random search is very powerful

• report in tables. with IQR 🅰 change figurtes into table with iqr

• the abstract is misleading, as it gives the impression that the paper targets the problem of selecting a suitable approach to solve a given SE problem. Instead, the paper selects the suitable hyper-parameter optimizer for a given approach (random forest in the paper);

  • framing

• R3: frameing

  • focus paper on the problems we are dealing with