Construction Dispute Potentials: Mechanism versus Empiricism in Artificial Neural Networks
Round 1
Reviewer 1 Report
These are my observations,
Ø There is need of strong literature support with justification of novelty of work.
Ø Data Validation part is not clear
Ø General discussion is more not linking each other test results
Ø Samples collection details are required for analysis- submitted / Received / Accepted level / Rejected.
Ø Source of sample / expertise levels.
Ø There are lack of recent literature survey. I have seen last 10 years only 2 or 3 papers are cited. Must be cite recent papers.
Ø Conclusion can be rewrite from results concisely not like general discussion. Its too lengthy.
Author Response
Dear Reviewer:
The authors wish to thank reviewer 1 for his/her time and effort in reviewing our manuscript. We have marked the changes in red in the manuscript. We hope the changes listed have made the manuscript suitable for publication and we look forward to your response.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Authors try to develop a way of artificial intelligence to deal with construction dispute. This is an encouraged idea. However (1) It is not clear for the motivation and goal. It seems that the introduction of intelligent algorithms is only to adapt to the current technology hotspot due to lack of description of relevance and the characteristics of the problem. (2) These two methods are the basic algorithms that widely applied. I can't see the innovation of this paper (3) The belief network and MLP is used to obtain the unknown inherent hidden laws by training a large number of samples. Why do author still chose BBN because it has been studied through probability theory based on the factor correlation analysis? (4) The MLP that focus on the fitting and recursion problem of continuous system will need high quality samples as training. It is hardly to obtain and quantify all construction disputes.
Author Response
Dear Reviewer:
The authors wish to thank reviewer 2 for your time and effort in reviewing our manuscript. We have marked the changes in red in the manuscript. We hope the changes listed have made the manuscript suitable for publication and we look forward to your response.
The authors
Author Response File: Author Response.pdf
Reviewer 3 Report
There are several concerns:
1-There are various studies that utilized ANN-based predictive models for construction litigation with many attributes (over 50) missed by the article,
2-There are various studies that utilized CBR-based predictive models for construction litigation with many attributes missed by the article,
3-The attribute selection process depends on a subjective process depends on the experts; the idea should be not to reduce the number of attributes but to utilize as many as because the disputes are complicated and cannot be described by limited attributes
4- This study should be tested with actual disputes so we can see the advantages of the study over the previous ones (20 years ago, it was done).
5-There is a need for a section for addition to BOK
Author Response
Dear Reviewer:
The authors wish to thank reviewer 3 for your time and effort in reviewing our manuscript. We have marked the changes in red in the manuscript. We hope the changes listed have made the manuscript suitable for publication and we look forward to your response.
The authors
Author Response File: Author Response.pdf
Reviewer 4 Report
This paper is investigating if empirical models aided by machine learning were able to bypass the necessity of understanding the formation mechanism. Contractual disputes in construction projects are taken as an example. Comparable machine-learning-aided mechanistic models and machine-learning-aided empirical models were established and their efficiencies were examined in different scenarios.
The paper is interesting and well-written.
Suggestion for authors is to add a Discussion section to compare the obtained outcomes and contributions with previous studies.
Author Response
Dear Reviewer:
The authors wish to thank Reviewer 4 for your time and effort in reviewing our manuscript. We have marked the changes in red in the manuscript. We hope the changes listed have made the manuscript suitable for publication and we look forward to your response.
The authors
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Although the authors add some literatures to indicate motivation the core problem of innovation and feasibility has not been solved. The neural networks are data-driven methods based on large amounts of data that indicate certain unknown laws. For this paper authors obtained the data by questionnaire which is usually suitable for statistical theory. Just from the sample, different from the process system strictly governed by internal and external laws, construction disputes are more complexity and specific. Hundreds of samples are too small which will lead to a lack of generalizability. In fact a construction dispute is not just a technical issue of fitting by historical experience that is a way out. It is difficult to be convinced by the result of neural network fitting (Even get a higher sample fitting accuracy than empiricism). It is suggested to seek some methods of objective evaluation such as game theory, key indicators evaluation, analytic hierarchy process and so on.
Author Response
Dear Reviewer,
Thank you for your comments. The authors have responded to the comments in the PDF file attached. Please kindly check. The authors appreciate your time and guidance in improving the manuscript.
Authors
Author Response File: Author Response.pdf
Reviewer 3 Report
Thank you for incorporating the changes. To improve the paper further, can you please explore articles by David Arditi starting in the 1997s. There are 6 or 7 articles specifically working on AI and Construction Litigation. Any paper exploring the same area must go through these articles. He has used both CBR and ANN, so we would love to see how your research further adds to BOK.
Author Response
Dear Reviewer,
Thank you for your comments. The authors have responded to the comments in the PDF file attached. Please kindly check. The authors appreciate your time and guidance in improving the manuscript.
Authors
Author Response File: Author Response.pdf
