A Complete Software Stack for IoT Time-Series Analysis that Combines Semantics and Machine Learning—Lessons Learned from the Dyversify Project

Round 1

Reviewer 1 Report

The paper devoted to results of research project Dyversify. Dyversify is an imec.icon project, where multi-disciplinary teams from academia and industry worked together to do demand-driven research. During the two-year project, seven different teams worked together and obtained results described in paper.

The full stack architecture is described. It combines data-driven and semantic methods, provides a high-level description of individual components, explains the design choices made, and discusses the problems encountered. The implementation is done using a Renson use case, in which air quality metrics from customer-owned ventilation units are used to detect events and anomalies.

The main results, which are described by the authors:

1. The fully functional hybrid microservice architecture for time series analysis is described;
2. The experience and problems encountered in the design and testing of the system are described.

Paper has some shortcomings that need to be corrected.

1. Background information can be shortened. Description of “Semantic Web” does not carry any semantic meaning, and is well known.
2. Related literature analysis is weak. Authors should concentrate and analyze defined researches dedicated to the topic.
3. Examples of sources 7-16 are useless. Authors just mentioned it without description of any outcomes.
4. Considering current pandemic of COVID-19 it is recommended to analyze researches on on-line data processing of coronavirus.
5. Architecture presented in Section 4 has weak scientific novelty. Authors should highlight the advantages of proposed architecture in comparison with known ones.
6. Novelty of the paper is weak and should be highlighted.
7. Authors should include discussion section to compare obtained with known researches.

The paper can be interesting to parties who are looking to build any type of hybrid data analysis pipeline from sensor to dashboard, but do not have relevant experience in developing such a system. It should be strengthened with scientific novelty.

In summarizing my comments, I recommend that the manuscript is accepted after major revision.

Author Response

Dear reviewer,

First and foremost, we would like to thank you for your adequate and helpful feedback. We have addressed your remarks and concerns, and made changes to the manuscript accordingly.

In the document (see attachments), we list our responses (including the specific change made) to each of your remarks. At the end of the document, you can find a full list of changes based on feedback from all the reviewers.

We hope the revised version can contribute to Innovations in Intelligent Machinery and Industry 4.0.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper present a proof-of-concept analysis platform, using a microservice
 architecture to ensure scalability and fault-tolerance that comprises time series ingestion,long term storage, data semantification, event detection using data-driven and semantic techniques, dynamic visualization, and user feedback. 

The paper is very well structured, being argued in detail the issue that is addressed. It also contains an analysis of specialized literature.  I have no comments to make.

 

Author Response

Dear reviewer,

We would like to thank you for the kind words and acknowledging the value of our work.

Kind regards,

The authors

Reviewer 3 Report

The authors in this paper are discussing an interesting topic, however:

1- The contribution of the project / the focus of the paper is not clear - the authors should mention the goals as well as the contribution clearly in points in the introduction.

2- The high-level architecture is very generic, and the authors in the different subsections of section 4 should clarify their design choices (e.g., the choice of microservices).

3- Anomaly/event detection is a very generic term, the authors should provide an algorithmic description of the selected event detection algorithm(s) and discuss their choices (e.g., given certain scenario/use case)

4- The paper is not showing any implementation (even for specific scenarios/use cases) or comparison with similar tools/platforms to prove the novelty and efficiency of the proposed algorithm.

Author Response

Dear reviewer,

First and foremost, we would like to thank you for your adequate and helpful feedback. We have addressed your remarks and concerns, and made changes to the manuscript accordingly.

In the document (see attachments), we list our responses (including the specific change made) to each of your remarks. At the end of the document, you can find a full list of changes based on feedback from all the reviewers.

We hope the revised version can contribute to Innovations in Intelligent Machinery and Industry 4.0.

Author Response File: Author Response.pdf

Reviewer 4 Report

The work presents a complex and modern architecture based on interdependent microservices for analyzing IoT data. The paper focuses on Cloud architecture and the management of acquired data. Data is thus processed to obtain a semantic labeling useful for the functions of the platform developed by the authors. It is interesting how authors analyze the possible solutions available for the various levels of their architecture compared to the alternatives on the market. However, given the complexity of the architecture, the work turns out to be quite long in reading without giving enough detail. In particular, the constant references to other works of the same group of research stand out (at least 9 ref. counted). Probably a descriptive figure of the microservices architecture and their interconnection could help. Furthermore, what is described in figure 2-Scalability through group based partition assignment in Kafka  is not very clear.

During text review, HVAC missing acronymous was founded.

Author Response

Dear reviewer,

First and foremost, we would like to thank you for your adequate and helpful feedback. We have addressed your remarks and concerns, and made changes to the manuscript accordingly.

In the document (see attachments), we list our responses (including the specific change made) to each of your remarks. At the end of the document, you can find a full list of changes based on feedback from all the reviewers.

We hope the revised version can contribute to Innovations in Intelligent Machinery and Industry 4.0.

Author Response File: Author Response.pdf

Reviewer 5 Report

Authors thoroughly describe an architecture of a system for time series analysis in the context of IoT devices. 

Each component is characterized and described in a proper and extensive manner, including details about specific technologies and the advantages and limitations they provide to the system. A use case is introduced, and related data and devices taken for explaining and exemplifying the components of the architecture.

The information/results presented in Section 5 seems interesting and relevant. However, even though there is a clear explanation of what was done and why it was done, it is hard to observe what was obtained and how it was measured and compared to make decisions on technologies and techniques. For instance, when describing the scalability requirements, phrases as "short startup", "fast shutdown", "exceptionally long" are not measurable. 

I recommend revising how results are presented and describe quantitatively the performed tests, data, number of events, performance of classifiers, etc.

Author Response

Dear reviewer,

First and foremost, we would like to thank you for your adequate and helpful feedback. We have addressed your remarks and concerns, and made changes to the manuscript accordingly.

In the document (see attachments), we list our responses (including the specific change made) to each of your remarks. At the end of the document, you can find a full list of changes based on feedback from all the reviewers.

We hope the revised version can contribute to Innovations in Intelligent Machinery and Industry 4.0.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Thanks for the authors for considering my comments. In my opinion, the paper now can be accepted.

Reviewer 3 Report

Thanks for addressing the reviewers' comments.

Reviewer 5 Report

Authors attended the recommendations from the first round of review.