Libraries need maps, too.
A run-through of my cartographic swag while working at Brooklyn Public Library
Maps are great. They really are. I'm honestly a strange candidate to dabble in cartography. Unlike most people, I'm more than happy to admit that I have the worst sense of direction out of anyone I know. Thanks, dad! Nevertheless, working at Brooklyn Public Library was an awesome opportunity to learn about the wonderful word of map making. The maps I made for my internship were for two main projects.
Five-Year Strategic Plan
The first project was the formation of the new five-year strategic plan. While I won't get into details, the crux of the plan revolved around creating "networks" of branches that shared characteristics in terms of geography, demographics and usage patterns. On my part, this involved a lot of work with the American Community Survey, looking at trends in cost of living, demographics and digital access ... and then putting it on a map.
Where should we build the next branch?
The other main project I worked on was helping determine the location of a new branch. The instructions were essentially, "We have a bunch of money. Where should we build a new branch?". Music to my ears. To answer this question, I relied heavily on patron-level data.
I first re-geocoded about 350k shoddily geocoded patrons' and then used R to calculate each patrons distance from the closest branch (the really beautiful map). This provided a great starting point, but as-the-crow-flies (Haversine) distance is only so meaningful.
To take things to the next level, a colleague and I (although mainly her) used the network analyst tool in ArcGIS to create 1/2 and 3/4-mile walking-distance radii around each branch. With these polygons in hand, I did some nifty geometric overlays in R to find which patrons fell within and outside the radii.
To top it off, I did some experiments with Census block groups to see if I could come up with some potential service areas. Block groups are the smallest geographic unit used by the Census Bureau - there are like 2,500 of them in Brooklyn.
I calculated the percentage of the population in each block group living outside the 1/2 mile radius, threw the data in a k-means clustering algorithm and iterated over it a gazillion times to find the sets of contiguous block groups that maximized the number of residents living outside the walking-distance radii. Below you can find an example of a Tableau dashboard I created to assist in the decision-making process.
It turns out that only about 5% of the population in Brookyln lives more from three-quarters of a mile from the closest library branch. So, what's your excuse?
Performing location-based video queries
A simple technique for searching YouTube’s vast catalog
A (fairly) simple technique for using Google’s kinda-sorta-really confusing Speech Recognition API
People, Politics and YouTube in Latin America
What do Latin Americans talk about in the comments section of YouTube? Is it possible to use YouTube comments to model and track public opinion on political issues?