Time Series Clustering Model based on DTW for Classifying Car Parks

Round 1

Reviewer 1 Report

The main issue and the novelty of the present paper is not clear. In fact, it is difficult to understand.

For instance, the definition of the main issue is barely defined in the first paragraph of the Introduction section. Also, it lacks in bibliographically defining how relevant and what is the contribution of the authors to this matter. Besides, some recent and important references are missing.

The writing is not adequate for publishing, since there are many typos in the whole paper and some words, such as “obvious“ is being repeated many times.

I will point only a few typos, but I strongly suggest a revision in the english writing. Please, check the lines 10, 11, 14, 19, 23, 33, 42, 121, 154, 165, 167, 178, 212, 236. There are many others, please check all of them.

The presented charts in Figures 1, 2, 3 and 4 are very simple and in poor quality, which difficult for the reader to visualize what the charts are proposing. I strongly suggest changing to vector charts using the EPS extension.

In the Data and Methodology section, the authors refer to 29 datasets and then changes to 27 because of missing values in two of them, since this information does not interfere in the remainingwork, there is no need to cite 29.

There are too short paragraphs that could be unified. For example, the paragraphs in Section 2.2.

Also, I had the impression that the font size changed along with the paragraphs.

In the Section 2.3, the authors presented results in a methodology section. Here I must point out some issues:

As the authors propose and defined three different methods to analyze the time series, it seems inappropriate that when it comes to applying them in data they propose a different analysis with different types of sets (that were not defined earlier) to pick only one of the methods. Since the objective of the paper seems me to be an analysis and comparison of the three different methods (as the authors wrote in Abstract), we have an incomplete work if all the proposed methods are not applied to the studied time series.

I suggest the application of all methods to the studied time series and comparisons of the results. Another option would be to re-arrange the paper to focus only on the used method (DTW+DBPAM) but that would require a better explanation of the model and a deeper literature review.

Concerning of results, the authors do not quantify the measurements. Therefore, it is difficult to visualize the pointed conclusions by the authors.

Author Response

We have uploaded a file for providing a point-by-point response to the reviewer’s comments.

Author Response File: Author Response.docx

Reviewer 2 Report

There are two focuses of the paper: (1) propose a time series clustering method namely Density-Based of Partition Around Medoids (DBPAM) clustering and (2) adoption of time series clustering to classify car parks based on the occupancy rate. 

First, the proposed DBPAM is mainly focused on the randomness and uncertainty of the original PAM at the initial stage. However, the proposed method seems computational intensive and only work for small dataset. The authors should evaluate the proposed method in term of time complexity in large datasets.

Second, the proposed framework of the application on car park clustering is simple. The result of the evaluation is trivial and can be observed easily without using the proposed framework if only 27 car parks are under consideration.

Author Response

We have uploaded a file for providing a point-by-point response to the reviewer’s comments.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The comments were addressed and answered by the authors. In my opinion, the paper can be published.