For a live demonstration of how the main algorithm works, please visit www.tearlant.com/mason-dixon
The demonstration should run out of the box in Python 3.10. To install all dependencies, at the command line, type
python -m pip install -r requirements.txt
Once the environment is properly set up, the demo should run on port 8888.
Welcome to the prototype of MasonDixon.
In all fields of data visualization, a challenge is to avoid plots that are misleading but technically accurate. This algorithm grapples with one such challenge in geospatial data analysis.
When colouring a map according to a variable (for example, per capita income or the average cost of a hotel per region), it is not trivial to divide a map optimally. On the one hand, we do not want too many large regions without cities. Conversely, if too many cities or population centres are bunched into a single region, it can skew statistics and make the map misleading.
MasonDixon is a geographic engine that combines geospatial, mathematical, and statistical methods to divide maps in an intelligent way.
To demonstrate the algorithm, a simple app has been built. The prototype is hosted at www.tearlant.com/mason-dixon
After the data loads, the map should be displayed with an overlay coloured according to the dataset:
By panning or zooming in/out, the map will dynamically redraw regions and re-visualize the data, based on which cities are now in the current field of view:
The app generates a set of random municipal data for all cities in the Natural Earth data set. This could easily be replaced with a numerical variable, such as the cost of the average hotel room in the region. (Most travel APIs do not offer this data for free, so I defer the creation of a travel app to the future).
The app then colours the map's field of view in a way that accurately visualizes the data. When the user is zooming and panning on the map, regions are dynamically divided and coloured based on the current field of view. The main algorithm divides the map so that there are never too many population centres in a single region.
After the map has been divided, the colouring algorithm takes a list of (latitude, longitude) pairs, e.g. a list of global cities, hotels, offices, etc., each of which has an associated numerical value. An aggregating function (such as linear regression or population-weighting) is then used to determine one overall value for the region, which determines its colour.
In the demo, if the user clicks the "Update Data" button, the values in the database will be updated (just adding random perturbations to existing values), and the new values will propagate to the map. In particular, these values are updated asynchronously, so the update could be easily swapped out with any web API call. For example, MasonDixon could be used to build an app that dynamically scrapes the average fare to fly to an airport or the average price of a hotel room in a city.
As a side note, most base maps use WebMercator coordinates to properly render, but most commercial web APIs (for example for travel data) use WGS84 coordinates. As such, the server was designed to use WGS84 coordinates when retrieving data, and perform the appropriate calculations to convert back and forth to WebMercator when communicating with the low-level mathematical and mapping functions.
Natural Earth. Free vector and raster map data @ https://naturalearthdata.com Bokeh: Python library for interactive visualization @ https://bokeh.org/ GeoPandas: Python tools for geographic data @ https://github.com/Geopandas/Geopandas Tornado Web Server @ https://www.tornadoweb.org/ ColorCET: Good Colour Maps by Peter Kovesi @ https://colorcet.com/