Next Article in Journal
An Adaptive Hierarchical Hybrid Multicast Based on Information-Centric Networking
Previous Article in Journal
A Hybrid Antenna with Equal Beamwidth in Two Frequency Bands for Radar Applications
 
 
Article
Peer-Review Record

Deep Time-Series Clustering: A Review

Electronics 2021, 10(23), 3001; https://doi.org/10.3390/electronics10233001
by Ali Alqahtani 1,2,3,*, Mohammed Ali 2,3, Xianghua Xie 3 and Mark W. Jones 3
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Reviewer 3: Anonymous
Reviewer 4: Anonymous
Electronics 2021, 10(23), 3001; https://doi.org/10.3390/electronics10233001
Submission received: 19 October 2021 / Revised: 27 November 2021 / Accepted: 28 November 2021 / Published: 2 December 2021
(This article belongs to the Section Computer Science & Engineering)

Round 1

Reviewer 1 Report

The authors reviews several time-series clustering methods. However, a major revision is required. Major comments are listed as follows. 1. Although the manuscript is a review paper, the authors should define the research questions in the first section. The authors should explain which research questions could be solved in this study. 2. The authors should propose an original review framework. The proposed review framework could solve these research questions. 3. The authors should compare the proposed review framework with other review frameworks. The differences between the proposed review framework and other review frameworks should be presented. 4. The authors should give systematical reviews based on the proposed review framework. 5. In Section 3.1.1, the authors should focuses on time-series methods in this study. For instance, the authors should explain why the Euclidean distance could be used for time-series analysis. 6. In Section 3.1.2, the authors should focuses on time-series methods in this study. For instance, the authors should explain why the Principal Component Analysis (PCA) could be used for time-series analysis. 7. In Section 3.2, the authors should explain why these methods could be used for time-series analysis. 8. In Section 4.1.1, the authors should explain why deep auto-encoder (DAE) could be used for time-series analysis. 9. The authors should discuss temporal auto-encoder and spatio-temporal auto-encoder in Section 4. 10. The authors should discuss the advantages and limitations of techniques. Minor comments are listed as follows. 1. The authors should summarize the contributions of this study in the first section. 2. The limitation of the proposed method and future work should be discussed in the last section.

Author Response

Please see the  attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors have presented a comprehensive, detailed review of time-series data analysis, with emphasis on deep time-series clustering (DTSC), and a case study in the context of movement behaviour clustering utilizing the DeepCluster method. They have reviewed these works and identified state-of-the-art, as well as present an outlook on this important field of DTSC from five important perspectives.
Their work from a technical point of view is very good which is worth publishing. However, the paper needs proofreading as there are several grammatical problems in the paper.  The abstract is well written, topics have meaningful full insight. The paper is accepted for publication.

Author Response

We thank the reviewer for their comments and feedback that contributed to enhance the content and presentation of our submission. In the revised version, the paper was proofread by Professor Mark W. Jones.

Reviewer 3 Report

This is a nice review work on the methods of clustering of deep time series.

The main concern of this reviewer is the appropriateness of this manuscript for Electronics.

Irrespective if this, the manuscript is written very well and is recommended for the publication.

Author Response

We thank the reviewer for their comments.

Reviewer 4 Report

In this paper, the authors present a detailed review of time-series data analysis. Some points should be included within the manuscript in order to improve the publication.

  • The authors should describe in detail the evaluation procedure of cluster quality.
  • In the conclusions section, the authors summarize the main points of their study. The authors should explain the contribution of their results in comparison to the results of other researchers.
  • The description of similarity measures and features extraction (section 3.1) in the present form is relatively weak and should be strengthened with more details and justifications.
  • The authors should refer to the recent papers:
    • Suboh Alkhushayni, Taeyoung Choi, Du’a Alzaleq, Data Analysis using Representation Theory and Clustering Algorithms, WSEAS Transactions on Computers, Volume 19, 2020, pp. 310-320.
    • Jacek Batóg, Barbara Batóg, Synchronization of Business Cycles in the EU: Time Series Clustering, WSEAS Transactions on Business and Economics, Volume 16, 2019, pp. 298-305.
    • Theodor D. Popescu, Time Series Analysis for Assessing and Forecasting of Road Traffic Accidents - Case Studies, WSEAS Transactions on Mathematics, Volume 19, 2020, pp. 177-185.

Author Response

REVIEWER 3

Comments: The authors should describe in detail the evaluation procedure of cluster quality.

Response: Two evaluation approaches to compute the cluster quality are described and discussed in Section 4.3.2.: Accuracy (ACC) and Normalized Mutual Information (NMI), which distinguish the clustering results generated by our deep cluster method and the ground truth labels.

Comments: In the conclusions section, the authors summarize the main points of their study. The authors should explain the contribution of their results in comparison to the results of other researchers.

Response: Based on our finding in section 4.3.2. the results were promising, showing that the latent space encodes sufficient patterns to facilitate accurate clustering of movement behaviors. Our study has compared DCAE and DCE, and shown that the clustering performance is efficiently improve by replacing fully-connected layers with convolutional ones. The clustering algorithm also performs much better compared to the original space clustering.

Comments: The description of similarity measures and features extraction (section 3.1) in the present form is relatively weak and should be strengthened with more details and justifications.

Response: We have revised the section mentioned as per the reviewer’s suggestions.

Comments: The authors should refer to the recent papers:

  1. Suboh Alkhushayni, Taeyoung Choi, Du’a Alzaleq, Data Analysis using Representation Theory and Clustering Algorithms, WSEAS Transactions on Computers, Volume 19, 2020, pp. 310-320.
  2. Jacek Bat´og, Barbara Bat´og, Synchronization of Business Cycles in the EU: Time Series Clustering, WSEAS Transactions on Business and Economics, Volume 16, 2019, pp. 298-305.
  3.  Theodor D. Popescu, Time Series Analysis for Assessing and Forecasting of Road Traffic Accidents - Case Studies, WSEAS Transactions on Mathematics, Volume 19, 2020, pp. 177-185. 

Response: We included the listed papers [2,3] to its related Section 3.1. Similarity Measures and Feature Extraction and paper [1] has been added to Section 3.2.2. Hierarchical Methods as it looked at how to analyze homology cluster groups utilizing agglomerative hierarchical clustering algorithms and methods.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have addressed my review comments to revise the manuscript.

Author Response

We thank the reviewer for their comments and feedback.

Back to TopTop