Huanfa Chen

Huanfa Chen

he/him

Associate Professor in Spatial Data Science, UCL CASA

Example of clustering on latitude and longitude

less than 1 minute read

Published:

Updates 2025/05/30

This note is a list of DOs and DON’Ts for conducting (spatial) clustering analysis. The examples are reproductions of student submissions.

DON’Ts

Clustering on longitude and latitude of points (without using other features & no justification)

image-20240530111903273

What it is: This is the k-means clustering of crime incidents based on longitude and latitude, which leads to the selection of 10 boroughs for further analysis.

Comments: The student justify using these 10 boroughs by clustering latitude and longitude coordinates, which is obviously going to cluster neighbouring boroughs and is not valid evidence for high-crime regions. If the objective is to identify high-crime boroughs, using the crime density (number of crimes per square km or per 1000 residents) would be more appropriate than clustering.

Comments

Introduction to Overleaf

3 minute read

Published:

Updates 2025/02/11 TL;DR: Using overleaf is much better than Microsoft Word (or MS Word) for academic writing that involves lots of formulas, and is very convenient for collaborative writing. UCL provides free and life-long Overleaf license.

Problems of MS Word in academic writing

  1. There are obvious differences between the desktop software of MS Word and the MS Word 365 on a browser, especially the formulas and figures.
  2. MS Word would crash if the file contains more than 100 pages, such as PhD dissertations.
  3. It is very hard to maintain lots of mathematical formulas in MS Word.
  4. It is time-consuming to change the template of MS Word. For example, IEEE journals requires two-column layout in submissions, which is a nightmare for MS Word.
  5. There is always a conflict if multiple people are simultaneously editing a Word doc via OneDrive, Dropbox, etc.
  6. etc …

Benefits of using Overleaf

  1. It is totally web-based and no installation is required.
  2. Many universities (e.g. UCL) provide free and life-long license of Overleaf, which allows many collaborators for one project.
  3. It is very easy to change the template. For example, if my paper submission to Journal A is rejected and I want to submit it to Journal B, it may take 30 minutes to change the template on Overleaf and resubmit. If I use MS Word, it might take hours or days.
  4. It has a rich text mode (see link), which resembles MS Word interface.
  5. It has functionality of “Track Changes” (see link) and “Add comments” (see link).
  6. It can be integrated with Github, Zotero, etc.
  7. There are lots of template of Overleaf. You can use it to write papers, prepare slides (using Beamer), prepare and update CVs.

Problems of using Overleaf

  1. The learning curve of Overleaf is quite steep. However, it is worthy for a four-year PhD and you don’t need to learn everything about Overleaf in one go (think about how you learn Python).
  2. As Overleaf is on a browser and depends on their server, sometimes the connection is unstable (but this is quite rare!).
  3. Some people don’t use Overleaf or resist using Overleaf. If your coauthor is one of them, you can either teach them to use Overleaf or download the Overleaf paper as MS Word document and send it to them. The latter way is really a pain, as you will maintain two versions of a paper in two different formats.

Tutorials of Overleaf

  1. UCL webpage and overleaf course at UCL: see link.
  2. 30-minute tutorial from Overleaf (link). Please note that you don’t need to learn everything about overleaf and Latex in one go. You can download an Overleaf template and write your own draft using template, and gradually learn Overleaf techniques when needed.
  3. Overleaf templates (link). Here you can find templates for UCL dissertation, Springer Nature LaTeX Template, etc. You can find more templates from Google/Github/CASA folks.
Read more