COVID-19 Situation Awareness Dashboard

•4 min read
Project
Back

It's already been more than 6 months since COVID-19 upended the global community. While it's frustrating reality to live in, it also opened doors to many researchers/developers outside epidemiologies to hone their data analysis and visualization skills, including myself.

Motivation

Like many others, I, initially was content with the Johns Hopkins University's famous COVID-19 Dashboard (opens in a new tab). It was a go-to place for folks to get to know up-to-date global statistics of the pandemics.

However, I began to sense that the visualization was somewhat not up to my expectation as it used scatter plot for worldwide infection status. Due to the exploding number of global clusters, the scatter couldn't be distinguished from each other, the inherent drawback of scatterplot.

After a bit of research, I found out Choropleth would be a solution and set out to do it myself, but not in Dashboard fashion.

Development history (From Tableau to Streamlit, a Web-based Dashboard)

Tableau

Initially, my first choice was Tableau, as I learned it could be used for geographic programming during my last semester. It turned out, unfortanately, my academic license for the software had already expired as it only covers for one year.

Python (geopandas)

So, I turned to Python, which has been my favorite progamming language so far, and found out I could mimic Tableau by using package called geopandas (opens in a new tab). After a quite bit of time tinkering and trial-and-error with GIS shapefiles and charting, I could make it happen. In fact, installing the package in conda environment was quite a hassle. As I didn't want to make a new virtual environment, I had to downgrade conda environment after quite a bit of research.

R (ggplot2, leaflet)

One day, I stumbled upon how visually appealing charts could be made by ggplot2 in R. I was quite shocked that I could make charts beautifully which I couldn't do with seaborn or matplotlib in Python.

After completing a specialization in Coursera (opens in a new tab), I began converting the previous works into R, adding more visual elements, namely, Heatmap and Interactive Choropleth powered by Leaflet library. With this transformation, the focus somewhat shifted from Choropleth itself to Heatmap, which ggplot2 excelled at aesthetically.

Streamlit (pydeck, altair)

Finally, I shifted again into Web dashboard.

Initially, I leaned toward Dash (opens in a new tab), because that was the only tool I had heard until then. However, I stumbled into Streamlit (opens in a new tab), which provides more cleaner, beautiful with less amounts of codes.

The only catch was I should really make a virtual environment as it doesn't allow installation on conda and suggest using pip. But, the benefits seem overwhelming. Without any styling definitions, resulting dashboard is sleek and modern-look. Aside from the look, it also boasts various graphic packages including Pydeck, Altair among others.

Dashboard components

To make good use of web dashboard structure, I divided the window into sidebar and main section.

In the sidebar, the following choices can be made.

In the main section, the following information are displayed.

Underlying tech stacks

To make this dashboard, following relevant sets are used.

Snapshots

Main Landing Page (Summary)

Heatmap

Choropleth

Future Developments

While I'm proud of what I could have done by myself, I acknowledge that it's not enough. Possible improvements include the following.

CC BY-NC 4.0 © min park.RSS