Applying Association Rule Mining to Explore Unsafe Behaviors in the Indonesian Construction Industry

Round 1

Reviewer 1 Report

The authors want to find the associations among unsafe behaviors and other construction attributes by evaluating the pattern of unsafe behaviors. However, the draft cannot be accepted in the current version.

(1) The association rule mining (ARM) method is adopted in the draft. However, why do you use this method but not other methods? The advantages of ARM should be distilled.

(2) Whether the literature review is comprehensive?

(3) The study may be too simplistic. What models or methods did the study propose? What are the specific reference values for safety management and safety performance?

(4) The innovation of this paper needs further refining?

(5) Figures 2-3 is not clear?

(6) Moderate English changes required.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors tried to uncover unsafe mining behaviors by using the data mining method with the self-designed questionnaire. The main idea of this work is to provide a data-mining scheme in the application of mining behaviors.

1. The contribution and novelty of this work are limited.  All methods used in this work are proposed by other people. The authors do not provide any improvements or development to the method. 

2. The section of the questionnaire design is a bit confusingly written, making it not easy to follow. How many samples and features were involved in this questionnaire? What does the statistical distribution of the collected data look like? Also, whether the collected data include some unbalanced or biased data? Have the authors explored the mean, and variance of the collected data? After looking through this section, I would suggest the authors do major revisions to the questionnaire.

3. The experimental section is also not clearly designed. How did the authors do the data mining? Have the authors trained the model before the testing? How did the author split the data set into training and testing? Have the authors done the cross-validation?  

4. The values in the tables are not clear? Do they represent the average result? It would be beneficial to report the values as the average with its standard deviations. 

5. Due to the no free lunch theorem in the machine learning area, it is important to compare the results of the proposed method against other competitors. Also, the method used in this work is a bit old.  To date, there are many learning-based methods the authors can explore. 

6. Since the authors proposed to explore and discover hidden patterns in the data, the impact of noise or outliers can be a great concern, which can potentially undermine the performance of most data mining methods. How did the authors deal with this problem? 

7. The introduction of this paper needs minor improvements as some references are a bit old. Like lines 42-46, it would be beneficial to introduce some SOTA methods in data mining. The venues can refer to ICDM, KDD, TKDD, SIAM-DATA MINING, etc. Also, the methods for exploring hidden patterns should involve the most classical method like PCA, NMF, manifold learning,  and learning-based methods like Autoencoder, please consider involving the following references:  

Vahdat, Arash, and Jan Kautz. "NVAE: A deep hierarchical variational autoencoder." Advances in neural information processing systems 33 (2020): 19667-19679.

Li, Xiangyu, and Hua Wang. "Adaptive Principal Component Analysis." Proceedings of the 2022 SIAM International Conference on Data Mining (SDM). Society for Industrial and Applied Mathematics, 2022.

Li, Bo, Yan-Rui Li, and Xiao-Long Zhang. "A survey on Laplacian eigenmaps based manifold learning methods." Neurocomputing 335 (2019): 336-351.

Liu, K., Li, X., Zhu, Z., Brand, L., & Wang, H. (2021). Factor-bounded nonnegative matrix factorization. ACM Transactions on Knowledge Discovery from Data (TKDD), 15(6), 1-18.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

1.      What is main advantage of the proposed model? What is main contribution to the field form your side? Make better presentation of the idea in relation to standard ideas and lates advances in the field.

2.      The research gaps in the abstract and introduction should be clearly expressed. Please rewrite this part.

3.      The authors must clearly explain the difference(s) between the proposed method and similar works in the introduction. The authors should further highlight the manuscript's innovations and contributions.

4.  One key background of this manuscript is the association rule mining and user behaviour analysis. Thus, the Introduction and/or related work section could be extended and incorporates additional discussions on the topics of user’s behaviour, e.g.,https://doi.org/10.1080/02564602.2014.968224,https://doi.org/10.1504/IJDS.2018.090623, http://dx.doi.org/10.1633/JISTaP.2014.2.2.3. This could set the scene and background for the subsequent discussions in this manuscript.

5. The authors may consider comparing the proposed method with more existing algorithms.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

In this paper, the author investigates many cases of unsafe behavior, it’s meaningful, but some questions need to be addressed:

1. The first occurrence of abbreviations (ARM) must give complete spelling in the abstract.

2. The literature review should address the existing problems, which will be solved in this paper.

3. Whether the weight of the influence factors should be considered?

4. It is suggested to give some statistical characteristics of the collected data.

5. It’s suggested that the authors refer to relevant literature to increase the reliability of the results in the paper. Three-dimensional discontinuous deformation analysis of failure mechanisms and movement characteristics of slope rockfalls. 

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

The novelty and contribution of this work are still limited as the method used (ARM) has been proposed many years ago. In the field of data mining, there are many state-of-the-art methods available that can outperform ARM. It should be noted that ARM is a method rather than an algorithm, and the terminology of algorithm is generally associated with optimizations that assist the objective function in finding a solution.

The authors acknowledge that the dataset may contain some biases. While it is common to encounter biases in real-world datasets, it is important to demonstrate how data mining methods can address this challenge. However, the authors have not presented the statistical characteristics of the dataset, nor have they described how they processed the biased data before the training process.

The impact of hyperparameters in data mining or machine learning cannot be underestimated, as they can significantly affect the final predictions and the reproducibility of results. However, the paper lacks clarity on how the authors tuned these hyperparameters.

It also appears that the authors have not adequately considered and addressed the effects of noise and outliers in this revision.

Overall, there are concerns about the findings in this manuscript as the experiments conducted may not be sufficient to support the proposed conclusion. Additionally, the experimental process lacks certain necessary steps. As a result, the quality and soundness of this paper are below the standard of this journal. Therefore, it may not be possible to accept this manuscript in its current version.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

Comparative study not found.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 3

Reviewer 2 Report

Although this revision is still below the acceptance criteria, particularly in the experiment section which lacks soundness and clarification, I have decided to accept the paper pending minor revisions. I still expect the authors to carefully address my previous concerns, as well as any additional comments provided during the review process. It is important that the authors address these concerns to ensure the validity and reliability of the study, and to provide readers with a clear and comprehensive understanding of the algorithm and its potential applications.

Author Response

Dear Reviewer,

Thank you for your valuable feedback on our manuscript. We are grateful for your input and have carefully considered your suggestions. We have added a new section that discusses the reliability of our findings. We have carefully selected a suitable method to evaluate reliability and have provided a detailed explanation of the approach in the revised manuscript.

Once again, we appreciate your insightful comments and feedback, and we hope that our revised manuscript meets your expectations.

Sincerely,

Authors