Let’s draw some polygons!

2 Toronto Header

KEY QUESTION: Given an Esri Shapefile, how can I visualize the coordinates as polygons?

DESIRED OUTPUT: A map of polygons

CATEGORY OF ANALYSIS: Geospatial analysis, data visualization

LEVEL OF DIFFICULTY: Easy

ESTIMATED SCRIPT TIME: less than 4 hours

DATA SOURCE: City of Toronto, City Wards, May 2010 (MTM 3 Degree Zone 10, NAD27)

CONTEXT AND DESCRIPTION: You have an Esri shapefile, which usually represents boundaries of some sort. Maybe later you want to see which points lie inside or outside of the polygons. Maybe later you want to make a chloropleth / heat map.

Before doing any of that, we need to first plot the boundaries. In this example, we use the City of Toronto’s open dataset of city wards. We plot the coordinates as polygons to see what the boundaries look like.

VALUE ADDED: This particular script and this particular data set are both very simple, and do not lend themselves to a lot of analysis. This is however the foundational script and building block for more complex analyses to be explored later on. Coupled with other data sets and additional functions, a lot more powerful and insightful analyses can be made later on.

In general, geospatial analysis offers many benefits, including a basic plot of boundaries. Many technical and non-technical people can interpret and digest the information on a map very easily. Using polygons can provide many insights. Maybe you have a boundary polygon and you have a list of addresses, and you need to see which addresses lie inside or outside of the boundary of interest. If you make a chloropleth/heat map, you can see the distribution and the change in your polygon in a very colourful manner.

METHODOLOGY: Using R, I programmed a script using the ggplot2 and rgdal libraries. The key ideas are to transform the shapefile in a useable dataframe and then use ggplot and geom_polygon functions to visualize the boundaries.

Here is the script:

Figure 1: R script of geom_polygon script

R script polygon

And voila! Here is one of the outputted maps:

Figure 2: Map of Toronto’s Ward Boundaries

Polygon

Again, this is a very basic example of geospatial analysis: it plots boundary lines on a map. This R library package offers many features. You can play around with colours, dimensions, shapes, transparencies, legends, etc. Heat maps or “choropleth maps” are also very interesting to see as they as show gradients of concentrations.

What are your thoughts? Do you find business value in this kind of visualization? Is the R script easy-to-understand? What can be improved? Leave a comment!

 

SOURCES: Here is a list of sources I used for help and inspiration. They’re great for more detail and additional reading.

Let’s plot some points on maps!

1 Balls Header

KEY QUESTION: Given a list of addresses, how can I visualize them as points on a map?

DESIRED OUTPUT: A map of the addresses

CATEGORY OF ANALYSIS: Geospatial analysis, data visualization

LEVEL OF DIFFICULTY: Easy

ESTIMATED SCRIPT TIME: less than 4 hours

DATA SOURCE: City of Toronto, Cultural Spaces

 

CONTEXT AND DESCRIPTION: You have a list of addresses: Maybe it is a list of establishments in a city. Maybe it is a list of customers and their addresses. Maybe each address represents an event (e.g. crime incident) that occurred there. Now you want to see them on a map. Maybe you want to see if there are any patterns. Which areas have high/low concentrations of points? Are the points evenly distributed? Are the points sparsely distributed? Are the points distributed in clusters? If so, what is the shape of the clusters? Are the points clustered in groups? Are they clustered in lines?

In this example, we use the City of Toronto’s open dataset of cultural spaces in the city. We plot each address as a point on the map to visualize the distribution of points throughout the city.

VALUE ADDED: Geospatial analysis offers many benefits, including a basic plot of points. Many technical and non-technical people can interpret and digest the information on a map very easily. Patterns on the distribution and clustering of points can provide many insights. If you understand the distribution and clustering of points, you can use this information to answer many questions. Where are my customers coming from? How can position my advertising in the most optimal location? Which areas have too many buildings or not enough buildings?  Which areas experience a lot of crime?

METHODOLOGY: Using R/R Studio, I programmed a script using the classic ggmap and ggplot2 libraries. The key ideas are to use the geocode function to assign latitude and longitude values for each address and then use geom_point and ggmap function to visualize the information in an aesthetically pleasing way.

Here is the script:

Figure 1: R script of ggmap script

Rscript

And voila! Here is one of the outputted maps:

Figure 2: Map of Toronto’s Cultural Spaces

Rplot

 

This is a very basic example of geospatial analysis: it plots points on a map. This R library package offers many features. You can play around with colours, dimensions, shapes, transparencies, legends, etc. Heat maps or “choropleth maps” are also very interesting to see as they as show gradients of concentrations.

What are your thoughts? Do you find business value in this kind of visualization? Is the R script easy-to-understand? What can be improved? Leave a comment!

 

EDIT: 2017-04-13 – As pointed out by Stephan in the comments, Google has a limit on how much you are permitted to geocode each day. If you go over the limit, you will receive an “OVER_QUERY_LIMIT” error. I believe the threshold is 2,500 records. In the code, there is a “write.csv” line because it would be prudent to save the records that you already geocoded. Since geocoding is very time-consuming and you have a finite amount you can run, then we should avoid duplicating work. There is no point in geocoding twice! Each day, you (or a group of your colleagues) can run the script to geocode large amounts of records.

The .xlsx file from the Toronto website was saved as .csv in the beginning before reading into R.
The other warnings are from the fact that Google could not geocode those records.