Capturing and Characterizing Human Activities Using Building Locations in America

Round 1

Reviewer 1 Report

In this paper, authors state a study of how multi-source data from building repositories and online social networks can be use to infer the actual activity of citizens.

The paper is well written and clearly describe some technical concepts involved in their course of action. Furthermore, they correctly put forward the correlation between the OSN and the building data and, thus, the fact that they can be fused and merged. They actually use a multi-level clustering approach to infer natural cities and hotspots inside these cities.

However, the paper does not describe how such clustering information can be used to infer the activity of citizens. Authors do not  provide any information about this core part of their work attending to the paper's title. In its present form, this paper just provides a clustering mechanism to infer human aggregations.  I would suggest authors to provide a mechanism that actually infers the activity of citizens based on the collected data with more semantic meaning. Otherwise, this work does not infer the human activity but human location.

Finally, a more comprenhensive comparative about different spatial clustering algoritms in section 3.1 should be done. What about hierarchical clustering?

Other minor issues:

- What Twitter checkin are? Are they just tweets with embebed posts from LBSN (e.g. Foursquare)?

- How do they extract such Twitter dataset? From an ad-hoc crawler, public repository? Please describe.

Author Response

Dear Reviewer:

 

Thank you very much for your timing feedback. Your review opinions are very insightful and I have revised my manuscript accordingly. I have agreed with most of your opinions and I would like to clarify at some points. In the following pages, I would like to response your valuable comments point by point.

Author Response File: Author Response.docx

Reviewer 2 Report

This paper aims to capture and predict the human activities and retrieve spatiotemporal information by using the buildings locations at city and country scales. The outputs are compared with the users twitters to investigate the reliability of social media for human activities prediction. I think the paper is totally relevant to the journal and readers would enjoy reading this paper. 

General comments:

1) The abstract and conclusion sections need to be revised/rewrite. The interesting outcomes of the paper are not properly reflected in these two sections. 

2) Integrating the discussion and conclusion sections is recommended.

2) The future work and the challenges of the research seems to be missed.

Detailed comments:

Stating the specific laws such as Zipf's law in the abstract might be confusing for readers. The authors can use a bit general outcomes of the paper in the abstract section.

What is the definition of natural cities? 

Sematic segmentation? 

 The resolution of Table 2 needs to be improved.

It would be better to add legend to the figures instead of describing them as caption.

The resolution of figure 6 needs to be improved.

Using spatiotemporal instead of spatial-temporal is recommended.

Line 162: The left boundary seems to be orange not yellow.

What do the grey and blue boxes represent at figure 1? Inserting a legend is required.

Author Response

Dear Reviewer:

 

Thank you very much for your timing feedback. Your review opinions are very insightful and I have revised my manuscript accordingly. I have agreed with most of your opinions and I would like to clarify at some points. In the following pages, I would like to response your valuable comments point by point.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Authors have solved all my concerns about the proposal. I would accept the manuscript in its present form.