Semantic Prediction Assistant Approach applied to Energy Efficiency in Tertiary Buildings

In principle, I find this a good paper that deserves publication in the semantic web journal. The main purpose of building a semantic prediction assistant aiming at energy efficiency in tertiary buildings makes sense and is well implemented. I do appreciate the responses given on the several questions in the first round of reviews, as I appreciate actions taken to change the manuscript. I think that the manuscript improved. Yet, I still would like to pass a number of comments:

1. Section 2.3 indicates that "in a BIM model, static information can of a building element can be queried, such as door: its material, when it was installed, or even the changes the door received until date." This is somewhat optimistic. Please make this less strong. 'When it was installed' might be by complete luck available, but a 'list of the changes' it has gone through, that will very seldom be included. This kind of information usually only comes up in the Facility Management (FM) phase (after construction), and they seldom use BIM models (mainly importing/exporting in the outset). They use FMIS systems.

2. Top left of Figure 2 should likely mention S4BLDG instead of BIM4EEPSA, as this rectangle only shows the top level concepts in SAREF, and not your extension. I still find it a pity that you reside to an appliances ontology, just to get a representation of a building structure. BOT does align with SAREF (https://github.com/w3c-lbd-cg/bot), so the argument hardly holds (clean and simple; integration with SAREF and IFC; documentation). You can't get it simpler than BOT, I would say; and the other two points are available via the SAREF alignment. Please at least change the URI to https://w3id.org/bot#. Furthermore, that W3C LBD community group is aiming to produce PRODUCT ontologies that extend from a bot:Element towards more specific elements (https://github.com/w3c-lbd-cg/product), as you actually also do in this paper (Door, Wall, Window, and so forth). In that sense, your efforts align quite well with that work.

minor text edits:
- section 3.1: more heterogeneous, encompassing
- section 3.1: and lodgings [72]. Therefore, the
- section 3.1: for automatically extracting proper
- section 3.2: thus provide them with semantic => rewrite: semantics? semantic annotations?
- section 3.4.1: no nearby sensors to compare its similarity => rewrite 'its'
- section 3.4.1: [The] data analyst is then assisted to make sure
- section 3.4.1: please reread this section, there are quite a few odd sentence structures still in there.
- section 3.5: in a form [that] data mining algorithms can accept
- section 3.5: focus on (instead of 'focus in')
- section 3.5: ETL (Extract, Transform, Load)
- section 3.5: [The] data analyst may have
- section 4: review the last two sentences of the first paragraph. They are grammatically incorrect (missing verb - verb entirely at the end of sentence).
- section 4.1: data analyst (instead of 'data-analyst')
- section 4.1: data source (instead of 'data-source')
- section 4.1: 31 January (instead of 1 January? - twice)
- section 4.3: Anyway, it will be [the decision of the data analyst] whether to incorporate
- section 4.3: Thanks to the knowlede-based [...] direct sun radiation. => reformulate sentence into proper English
- section 5.2: currently, only the SemOD Method for detect temperature [...] is implemented => reformulate sentence into proper English
- section 5.2: This is why, research -> This is why research

The paper tackles the interesting problem of using a semantic prediction assistant within the area if energy efficiency of buildings. To the best of my understanding the core of the problem is trying to help data analysts in the knowledge discover process of building sensor data. The approach proposed is the Energy Efficiency Prediction Semantic Assistant (EEPA) that uses ontologies to support the data analyst in knowledge discovery to support the definition of HVAC strategies in a semi-automatic way.

Overall the paper was enjoyable to read, but I did have some challenges understanding key parts and it required multiple readings. I have some suggestions that I think may improve the paper:

=Section 1=
- In the introduction is [68] the correct reference for the first paragraph?

- The contribution of the paper was unclear to me in the introduction. You say EEPSA uses SWT in KDD and leverages expert knowledge, but the exact scientific over existing works is unclear. SWT have been used in KDD before, what is the specific contribution here.

- The exact problem tackled is also unclear in the introduction. Again, SWT in KDD is very high-level. I would expand the description to be more specific.

=Section 2=
- In general section 2 is OK in terms of the work covered. To me it read as a justification of the need for the EESPA ontologies and as a motivation for the reuse of concepts. Reusing existing ontologies within EESPA is a welcome choice.

- However, I found it very difficult to understand the analysis of the related work as the problem the paper tackles is not clearly stated at this point. I did not know the detailed motivation for EESPA and what it was trying to solve. HVAC control strategies and the challenges associated with then would be useful for the reader to know.

- I would recommend introduction a motivation scenario and a requirements analysis before section 2. You start to do this in section 3, but it would help the reader if this was introduced sooner.

- Section 2.2 “SWT for KDD” introduces a lot of related concepts including Linked Data, Open Data, and data mining tools. I was unsure what the exact message of this section was.

- After reading section 2 it was unclear exactly that the gap in the state of the art was, and what the contribution of EEPSA is. The section serves as a good justification for the design of the EEPSA ontologies, but needs to more clearly define the contribution of the work.

=Section 3=
- The motivations of problem and requirements are briefly introduced at the start of section 3. I would encourage you to move this discussion to the start of the paper. I would also suggest you extend the discussion to be much more specific on the requirements of the problem and the contribution of EEPSA. Currently, this discussion is a little too general.

- The design of the two presented ontologies look good. I applaud your efforts to reuse existing works. Well done.

- It may help to improve the description if you first gave a high-level description of the different ontologies involved in EEPSA, and detail how they meet the identified requirements.

- In general section 3 details a number of contributions (i.e. outlier detection support) and possible future contributions of EEPS in the different phases of KDD. I found this a little difficult to follow. I would encourage a focus on the core contributions of this paper.

- Due to the above observation I am not sure if structuring section 3 along the KDD steps make the most sense.

=Section 4=
- Section 4 presents a nice description of a real-world evaluation. The setup is very interesting and helped to understand the problem tackled in the paper. Perhaps this should be used to support the motivation at the start of the paper.

- In terms of the evaluation I feel it needs to focus a little more on supporting the contribution of the paper. Make clear how it validates the EEPSA and the ontologies defined.

Overall this is an interesting paper which can be improve significantly through a restructuring to improve the communication of the problem and the contribution of the work.

I acknowledge this revised version of the paper takes into account the reviewers comments, is better written, more concise, more clear.
The new version of the EEPSA ontology is now well described in section 3.1, which is an important addition to the previous version.
I see no major new issue in this paper now, and recommend if for publication.