CODE: A Moving-Window-Based Framework for Detecting Concept Drift in Software Defect Prediction
Round 1
Reviewer 1 Report
- In section 1, you should describe the background of your research before you mention the limitation of the existing solutions.
- Abbreviations of words (CD, CV,…) defined as duplicates and correct them.
- There are research works including CNN for detecting software defect prediction. You should describe why you choose traditional machine learning algorithms in your experiment.
-
Please add the precise limitations and future expansions of the study to the conclusion(section 9)
Author Response
Reviewer 1
Comments and Suggestions for Authors
1. In section 1, you should describe the background of your research before you mention the limitation of the existing solutions.
Response:
We are thankful to the reviewer for the valuable comment, as this will provide a better readability to the manuscript, we have now described in more detail about the research’s background in section 1, page 1-2, lines 32 - 39, as follows:
The software entities (such as files, packages, and functions) in a software system that are prone to defects are identified through models for software defect prediction (SDP). An accurate prediction model assists developers in concentrating on the anticipated flaws and efficiently using their time and effort. Prediction models collect knowledge about a software project by studying previous software information, and they can then predict whether or not instances introduced in the future will be flawed. To build a reliable and high-quality software system, there are many existing software defect prediction (SDP) models (e.g., [ 4–8]) that use machine learning to predict faulty modules in software system.
2. Abbreviations of words (CD, CV,…) defined as duplicates and correct them.
Response:
We are again thankful to the reviewer for pointing out this mistake in the manuscript. The correction has been made in the revised manuscript on page 1, lines 1, 8 and page 2 line 65, as follows:
- Concept Drift (CD)
- Cross-Version (CV)
3. There are research works including CNN for detecting software defect prediction. You should describe why you choose traditional machine learning algorithms in your experiment.
Response:
We are thankful to the reviewer for asking about the selection of conventional machine learning algorithms for classification. The primary purpose of the study is to determine the influence of class imbalance and concept drift both on the classification accuracy, and the conventional machine learning algorithms can serve the purpose therefore we didn’t utilize the CNN.
4. Please add the precise limitations and future expansions of the study to the conclusion(section 9)
Response:
We are thankful to the reviewer for the valuable comment. The limitations and future work section were missing in the manuscript therefore we have added these details on page 18, lines 608-612, under section 9 as follows:
The study is performed specifically on software defect datasets; therefore, the proposed framework cannot be generalized to other applications. As future work, the proposed framework would be tested on more generalized datasets to increase the applicability of the framework. Moreover, as the proposed framework relies on the labeled data, this could limit it use in specific applications where the labeled data is not available.
Reviewer 2 Report
the article should be supplemented with a section indicating what the practical implications of the experimentation carried out are and what the future developments may be.
Author Response
Reviewer 2
the article should be supplemented with a section indicating what the practical implications of the experimentation carried out are and what the future developments may be.
Response:
We are grateful to the reviewer for the valuable comment. Based on the reviewer’s concern we have added the details related to practical implications on page 18, lines 600-607 as follows:
In practice, software quality practitioners are more interested in techniques that can effectively allocate testing resources. Thus, the knowledge of the stability of past prediction models is necessary for practitioners in such NSE. To that end, the CODE framework assess the robustness of class rebalancing techniques those help the practitioners in allotting the scarce testing resources, efficiently. From the experiment, the significant results demonstrate the impact of the proposed framework. Although this empirical study is conducted for CVDP, the proposed CODE framework could be employed for any defect datasets where the data appears over time and also based on the availability of labeled data.
The future developments and limitations of the proposed work are added in section 9 on page 18, lines 608-612, as follows:
The study is performed specifically on software defect datasets; therefore, the proposed framework cannot be generalized to other applications. As future work, the proposed framework would be tested on more generalized datasets to increase the applicability of the framework. Moreover, as the proposed framework relies on the labeled data, this could limit it use in specific applications where the labeled data is not available.
