Tri-Partition Alphabet-Based State Prediction for Multivariate Time-Series

Round 1

Reviewer 1 Report

The manuscript presents a proposal of a tri-partition alphabet-based state (tri-state) prediction method for symbolic multivariate time series. The work is intended to obtain richer semantics allowing similar or better performance. Despite the idea of the proposed method is perceived, it is not so easy to realize the main developments of the tri-state prediction method when compared to previous methods. Consider the following general suggestions:

1. It would be helpful to include a comparison with other methods regarding implementation aspects, e.g., conceptual design, data treatment, and prediction scheme.
2. The mathematical formulation of the method is deeply addressed, but it would be helpful to have some figures to illustrate its architecture.
3. The demonstration examples are presented very generally so that the real-life application of the method is hard to understand. It would be interesting to refer to the specific variables that are being predicted.

The manuscript is generally well-structured, but there are faults in the writing, and the location of tables and figures is not so fine. Specific comments to address are given below.

4. The Intro includes a very general state-of-the-art and then moves to a deep description of the method. A more comprehensive presentation of the problem should be made and, if possible, with reference to more specific examples.
5. The description of the method can start from a simpler description of prediction algorithms, e.g., with reference to the fundamentals of the k-NN method and including illustrations.
6. Please avoid writing in the first person (We). There are writing mistakes in the manuscript, e.g., ‘There are two things [that] need to be explained. Please perform a careful revision.
7. Line 130: Check that the last term in the first row of O10 is correct. Line 153: Explanation of PAA and SAX as being equal to True or False to fit the data is not so clear. Please improve.
8. Section 3: Avoid starting this section with ‘Figure 1 presents…’. Figure 1 and Table 5 are presented before they are referred to in the text.
9. Section 4, Intro: Provide a more explanatory introduction for this section, e.g., by moving the first sentence in 4.1 to above.
10. Section 4.2: How is the precision metric defined and calculated? Figure 3 can be placed immediately after Figure 3. When comparing the metrics in Figures 2 and 3, maybe it would be fairer to do it in relation to the same (r, c) combination. When predicting new data which are not in the database, which criterion to use to choose the r and c indices?
11. Section 4.3: Tables 6 and 7 should be placed at the beginning of the section, where they are referred to. Similar for Figures 4 and 5.
12. The Conclusion is mostly a statement of what was done. The main findings of the work should be highlighted.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Topic of the manuscript isn’t close to the fields of my expertise. However, I can recognize an extensive manuscript. It includes 66 references, but 9 of this references are presented at the international conferences and symposiums. In my opinion, some of these 9 references could be replaced by the references published in leading international journals.

In addition, the English language usage should be improved (maybe, by native speaker, if possible).

The appropriate overview of the existing literature is given into the Introduction of the manuscript, but descriptions of the considerations performed through the manuscript are so wide. In my opinion, the performed considerations should be briefly described in the Introduction, but widely in the corresponding sections.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Authors satisfied all my requirements from the first round of review process. On my suggestion, authors have reduced the total number of references (from 66 to 56) and have removed all conference papers from References. Besides, they have added new section 2 which includes some descriptions removed from the Introduction of the initial manuscript version.

The revised manuscript version is better than the initial one. Now, this version is almost ready for publication. Only, the English language usage could be additionally checked (I recommend, by the native speaker).

Author Response

Please see the attachment.

Author Response File: Author Response.pdf