Recognition of Intersection Traffic Regulations from Crowdsourced Data

Round 1

Reviewer 1 Report

This research was comprehensively conducted and reported. It clearly introduced the research question and the proposed methodology for recognizing intersection traffic regulations from spatial crowdsourced data. Overall, the manuscript is largely ready for publication. Below, I will list a few minor comments and suggestions:

(1)   I have found several typos and grammatical mistakes in the main text. For example, “their adoption the crowdsourcing scenario” in Lines 73-74, “where in feature vector information” in Line 152, “threfore this study” in Line 347, etc. Please do editorial checks carefully.

(2)    For readability, I would suggest the authors explicitly define/describe what an intersection arm is. The concept was not well known in the GIScience community. Is it directed? 

(3)   In Figure 4, the fourth movement pattern was not clearly illustrated. Essentially, a vehicle has to decelerate before each stop.

(4)   In Figure 6, I would suggest that authors use a clearer font type.

(5)   In Section 2.2.3, the rule of combing features from all the arms of the same junction should be explicitly mentioned. Is it concatenated in clockwise or anti-clockwise order?

(6)   In Table 4, the bold number was incorrectly labeled for the Static RF model of Chicago. Moreover, the statement in Line 441 “0.95 in Champaign and Hanover, and 0.82 in Chicago” was inconsistent with the information in Table 4. Similarly, in Line 452 “Similar results are observed for the RF classifier, except for the Hanover dataset” was also problematic.

(7)   In a sense, the proposed Hybrid model just outperformed the other models marginally (for example, about 1% in many cases). Meanwhile, the authors’ strategy of model tuning also demonstrated a 1% increase in accuracy. It is difficult to tell whether the hybrid model is better than the static and dynamic models. 

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper proposes a traffic intersection detection method based on crowdsourcing trajectory data. The authors designed their classification algorithms using Random Forest and Gradient Boost. Three datasets are tested to evaluate the suggested methods. The overall quality of this paper is high both in framework design and representation.  However, there are still some issues that should be addressed before the paper can be considered for publication. 

1 The results of classification or detection are not compared with other existing studies. The author should give some information about the GPS trajectory-based traffic intersection detection and compare the results with existing work.

2 The quality of the GPS trajectory may affect the results of detection. From the given GPS trace figure, we can see many errors in the trajectory. How this will affect the final analysis results should be addressed. Furthermore, some trajectories that are against the traffic laws may also affect the classification results, which should be explained and analyzed.

3 It is also interesting to analyze classification accuracy considering the number of trajectories in different intersections. How many trajectories are required to identify the intersection information?

There are also some specific suggestions on the figures:

The font in Figures 5 and 6 should be improved.

Figure 9 is also difficult to read.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The authors have revised the paper according to the comments from viewers. It is suggested to publish this paper after minor revisions in languages and formatting. 

Author Response

Thank you very much for your review. The article has been checked and corrected from a native British speaker. The corrections are shown in blue in the text and at the end of the article there is also a list with all the changes. We believe that now the text contains no grammatical and syntactical errors.

Reviewer 3 Report

The authors have addressed all points raised in my review. I support the publication of the article.

Author Response

Thank you very much for your review. The article has been checked and corrected from a native British speaker. The changes in the original text are shown in blue. At the end of the article we have included a list with all the changes/correction of the article. We are now sure that there is  no grammatical, syntactical or any other kind of language error in the article.