Batman’s Trending: A Spatial Approach to Normalizing Google Trends and it’s Inferences on Geographic Data

With the growing fascination of map forums such as Map Porn on Reddit or Terrible Maps on Instagram, it’s evident that mapping has gone beyond the needs of the typical geographer. Open data is used every day and sometimes use to produce “cool” map that nevertheless, relays information. One of my obsessions is superheroes from the DC and Marvel universe. Comic books, movies and conventions have steadily increased and with Disney overseeing all Marvel movies this trend will not be dwindling anytime soon. Google Trends is one of the most useful, real time and accessible data resources that is freely available to anyone who has internet access. Google Trends is keyword-related data that uses the volume index as well as the amount of search engines within the geographical area. I wanted to explore this type of data to see its limitations, and its usefulness.

The limitation with Google Trends is the output data provided has already been normalized by Google themselves. Google’s normalization algorithm is not provided, instead, they state:

“While only a sample of Google searches are used in Google Trends, this is sufficient because we handle billions of searches per day. Providing access to the entire data set would be too large to process quickly.”

 Instead, results are normalized through the following process:

  1. Data points are divided by the total searches of the geographic location and time to compare relative popularity.* This eliminated locations with the most search volume being ranked highest.
  2. Results are scaled from 0 to 100 range with consideration to the topic’s amount to all searches on every topic generated.
  3. Regions that show the same search interest will not have the same total search volume.

*it should be noted that Google’s normalization method throughout time causes data to slightly changes depending on the day you gather the information despite the time frame you consistently use. To have a more accurate sets of data, it is best to gather all the data within the same week to avoid larger differences

The design of the Geovis project is that of a physical map and the aesthetic of an old fashion comic book pop art piece, that is simple, artistic and comprehensive of the data. 

Laser Cut

After speaking with the laser cutting company Hot Pot, three laser cut pieces were produced, one of the United States, one with all the States, and one with Alaska and Hawaii. The maps are on a 1:20,000,000 scale which is then be blown up to 2.5x its size to fit the boards. Each state will be covered with its trending superhero or villain from within the past year.

In order to create the laser cut template, I imported a shapefile of the US states into ArcPro and used Albers Equal Area Conic projection. A dissolve tool was then used on the US shapefile to create an outline of the continental United States. The maps were then exported as an SVG in order to create a vector file the company could use. Since there are many intricate details of the US states there were some errors from the SVG file ArcPro produce. To correct this the US state file was imported into Adobe Illustrator and traced to create the vectors.

Next, the superhero/villain data needed to be gathered. Below is a step by step process on how I normalize Google Tends Data and create a visual physical map that any fan would love to hang in their living room.

Google Trends Superhero Data

17 characters were chosen from both universes to populate 50 states. 20 male characters and 14 female characters that have had a major motion picture or television appearance between 2017 and 2019, which exception to staple characters and upcoming movies in 2020.

DC Superheroes and antiheroes
that have made an
appearance in 2014 -2019
Marvel Superheroes and
antiheroes that have made an
appearance in 2014-2019
Batman (Movie 2017) Iron Man (Movie 2019)
Superman (Movie 2017) Captain America (Movie 2019)
Joker (Movie 2019) Hulk (Movie 2019)
Aquaman (Movie 2018) Spiderman (Movie 2019)
Flash Barry Allen (TV show 2019/Justice League 2018) Black Panther (Movie 2019)
Shazam (Movie 2018) Thor (Movie 2019)
Cyborg (Movie 2017) Venom (Movie 2019)
Green Lantern *staple in the DC universe Deadpool (Movie 2019)
Green Arrow (TV show 2019) Groot (Movie 2019)
Dick Grayson (TV show 2019) Hawkeye (Movie 2019)
   
WOMAN WOMAN
Wonder woman (Justice League) Black Widow (Movie 2019)
Harley Quin (upcoming movie 2020) Captain Marvel (Movie 2019)
Catwoman *staple in the DC Universe Scarlet Witch (Movie 2019)
Raven (TV show 2019) Gamoura (Movie 2019)
Starfire (TV show 2019) Storm (Movie 2019)
Supergirl (TV Show 2019) Jean Grey (Movie 2019)
Bat Woman (TV show 2019) Nebula (Movie 2019)

Once all characters were established each character was imputed into Google Trends to find there rating. A year from September 30, 2019, was used as the time frame and a score 75 and above was recorded on an excel sheet in order to keep track of all the rankings. The Related Topic and Related Queries tables were used to make sure the term was relating to the character and not something else. If the character showed unrelated topics and queries because of the ambiguity of their name, they were kicked out. To keep universes equal in numbers, if one character was kicked out of the list another from the opposite universe needed to be kicked out as well. Venom and Storm were kicked out of Marvel and Shazam and Raver were kicked out of DC.

Once all the characters were recorded the next was to determine how a superhero/villain won a state. Since I wanted all characters to be represented on the map, I first ranked the character scores and then the state’s score.

If a character scored 100% for a state and no other score the character instantly won that state. These included Aquaman, Hawkeye, Green Lanter and Green Arrow.

Utah could not be won by one superhero alone because of the large amount of 100% scores. Instead, Utah would be split into 4 characters who score the most 100% scores.

Next, I looked at superheroes with the lowest number of total scores. If a superhero had a low total number of scores but only one 100% ranking score the superhero would win the state. If there were multiple 100% scores for a state (Alaska) those states needed to be ranked later on.

Once all the characters won over a state the next step was to rank the state scores. Whichever character held the highest score within that state they won that state.

Once all the scores were established the next step was to make the physical map

Making the Map

First thing to do was gather all necessary comic books. After attaining over 30 I had all my characters.

After receiving the laser cut of the map each character was cut out to fit the shape of the state and glued and sealed onto the continental US map. Since Alaska was too big to fit on to the same map as the continental US, it was given its own separate board along with Hawaii.

After all characters were glue down and sealed, fabric reflecting the original pop art colours was glued down to foam board before mounting the maps. This step was done for both the continental map and Alaska/Hawaii map.

Since some east coast states were too small to have characters directly represented on the stated, yellow fabric was used as filler. The east coast states where then outline and used an array of coloured strings and pins to connect pictures of the state characters to the shape of the states then to the geographic location.

Once all the states were now represented by characters from both universes a scale bar was added to both maps as final touches.

The maps turned out amazing and something truly artistic. I could not make this map alone and would like to thank my friends in the program Jeremy, Miranda and Fana for all their help and support, my friend Kevin who helped with all Adobe Illustrator needs, and my cat for being himself.