Regional Traffic Event Detection Using Data Crowdsourcing

Round 1

Reviewer 1 Report

Conclusively, my suggestion for this article is to accept it after a major revision. The main reason is strange results presented in this manuscript and lacking some critical content. I suggest the authors make the following revisions:

1. In this article, the authors concluded that a linear support vector machine algorithm outperformed a nonlinear one in classifying their data. This conclusion is strange. I suggest the authors list the kernel function and other necessary parameters adopted to implement the nonlinear support vector machine algorithm. Moreover, the authors may add a figure to illustrate that a linear support vector machine algorithm is more suitable for classifying their data than a nonlinear one.

2. The authors also concluded that the linear support vector machine algorithm outperformed a convolutional neural network in solving their problems. This conclusion is still strange. A convolutional neural network is a more complex classification algorithm than a linear support vector machine algorithm. I suggest the authors briefly describe how their convolutional neural network was constructed.

3. The authors have not provided a statistical description of the data employed in this article. Please add such a description for the acceptance of the revised manuscript.

In conclusion, I suggest the authors add sufficient figures or content to support their results.

 

Moderate editing of the English language is required. Some grammatical errors are found.

Author Response

Dear Reviewer,

I would really appreciate your good comments.

We did our best to reflect your comments.

Please refer to the attached files on the author's revision notes and the revised paper.

I hope you accept our paper for publication in Applied Sciences.

Many thanks.

 Jaesoo Yoo

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper develops a technique to detect regional traffic events using the so-called crow-sourcing method. It claims that the proposed method is valid and effective.

 

However, the paper does not describe where the crowd-sourcing posts are coming from, and the authors do not provide an example of the posts. On page 11, the authors mention that “The content indicates traffic conditions such as accidents, construction, etc., and consists of text.

”, if the post itself contains so much information, why wouldn’t we just utilize that information? We do not need all these fancy machine learning models to categorize them into different types and do complicated post-processing.

 

Also, from my personal experience, effective crowd-sourcing normally needs to rely on more rich data sources, like Twitter and Facebook, where millions of posts will be posted daily. The TBN Korea Traffic Broadcast, from my understanding, is not a platform where a lot of public users will post their stories. So the data effectiveness remains questionable.

 

In summary, I have doubts about the effectiveness and contribution of this study to detecting regional traffic.

Author Response

Dear Reviewer,

I would really appreciate your good comments.

We did our best to reflect your comments.

Please refer to the attached files on the author's revision notes and the revised paper.

I hope you accept our paper for publication in Applied Sciences.

Many thanks.

 Jaesoo Yoo

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

I can have no further questions. The review can end.

Minor editing of English language required.

Author Response

Dear Reviewer,

We would like to sincerely thank you for your attentive indications. We we have carried out editorial revisions through MDPI English proofreading service in order to reflect your comments on the Quality of English Language.  Please refer to the attached revised manuscript.

Many thanks.

Jaesoo Yoo

Author Response File: Author Response.docx

Reviewer 2 Report

Thanks for the response. Regarding the revision, I have two follow-up questions:

 

1. In the revised version, the authors have incorporated data from external sources such as Twitter and broadcast stations in accordance with the suggestions made in my initial review. However, the methodology employed for this data integration has not been adequately elucidated, nor have the authors provided information regarding the distribution of data points from each of these sources. Given that the inclusion of these additional data sources represents a nontrivial alteration, encompassing extensive data cleaning and preprocessing efforts, the current textual description alone fails to substantiate the efficacy of this modification. In order to address this concern, it is imperative that the authors furnish a detailed account of their data collection procedure, explicitly outlining the steps taken to integrate the new data sources and specifying the volume of data contributed by each source. Such comprehensive clarification is essential to validate the integrity and credibility of this new data augmentation process.

 

2. Regarding reviewer #1's comments about the superiority of Support Vector Machine (SVM) over the neural network, the response from the authors appears to be insufficient. In the revised manuscript, the authors state, "The results are due to the fact that the Linear Support Vector Machine Algorithm is better suited for both the characteristics of the crowdsourced data we used and the problem of text classification." However, this argument lacks clarity and supporting evidence. It is imperative that the authors provide a comprehensive explanation of the specific characteristics of the crowdsourced data employed in the study and why SVM is deemed more suitable for it. Without substantial justification, this statement risks being perceived as unsubstantiated. To enhance the validity of their claim, the authors should present a well-substantiated explanation supported by concrete evidence.

N/A

Author Response

Dear Reviewer,

 

We would like to sincerely thank you for your attentive indications and good comments. We partially rewrote and complemented our paper in order to reflect your comments. Please refer to the attached revision note for your second review comments. 

Many thanks.

 

Jaesoo Yoo

Author Response File: Author Response.docx