Visualising Google Trends results with R

I haven’t been blogging as much as I’d like to due to other commitments but I wanted to write a post before 2018 ends. This post is on plotting Google Trends results with R. If you’ve never heard of or used Google Trends, it’s fun! You can see how certain keywords have trended over the years and have the results broken down into regions and countries. In this post, I’ll use the gtrendsR package to perform Google Trends searches with R and show how to plot the data that is returned.

To get started, install the gtrendsR, dplyr, and ggplot2 packages (if you haven’t already).

install.packages("gtrendsR", "dplyr", "ggplot2", "maps")

# load packages
library(dplyr)
library(ggplot2)
library(gtrendsR)
library(maps)

The gtrends function performs the search. The function has several parameters; I’ve set geo to focus on the United States and time to include all results since the beginning of Google Trends (2004). I like the San Antonio Spurs, so that will be the keyword for this example.

res <- gtrends("san antonio spurs",
               geo = "US",
               time = "all")

names(res)
[1] "interest_over_time"  "interest_by_country" "interest_by_region"  "interest_by_dma"     "interest_by_city"   
[6] "related_topics"      "related_queries"    

The results are stored in a list that have self-explanatory names. I picked a keyword that should be region- and city-specific. Below I show the interest in “san antonio spurs” stratified by region, city, and DMA and sorted by the interest (highest to lowest).

res$interest_by_region %>%
  arrange(desc(hits)) %>%
  head(3)

    location hits           keyword geo gprop
1      Texas  100 san antonio spurs  US   web
2 New Mexico   26 san antonio spurs  US   web
3   Oklahoma   21 san antonio spurs  US   web

res$interest_by_city %>%
  arrange(desc(hits)) %>%
  head(3)

        location hits           keyword geo gprop
1    San Antonio  100 san antonio spurs  US   web
2        Schertz   91 san antonio spurs  US   web
3    Canyon Lake   75 san antonio spurs  US   web

res$interest_by_dma %>%
  arrange(desc(hits)) %>%
  head(3)

           location hits           keyword geo gprop
1    San Antonio TX  100 san antonio spurs  US   web
2         Laredo TX   68 san antonio spurs  US   web
3 Corpus Christi TX   54 san antonio spurs  US   web

I modified the plotting function from the gtrendsR package, so that the function just returns a ggplot object without plotting it, so that I can customise the plot.

plot.gtrends.silent <- function(x, ...) {
  df <- x$interest_over_time
  df$date <- as.Date(df$date)
  df$hits <- if(typeof(df$hits) == 'character'){
    as.numeric(gsub('<','',df$hits))
    } else {
    df$hits
    }

  df$legend <- paste(df$keyword, " (", df$geo, ")", sep = "")

  p <- ggplot(df, aes_string(x = "date", y = "hits", color = "legend")) +
    geom_line() +
    xlab("Date") +
    ylab("Search hits") +
    ggtitle("Interest over time") +
    theme_bw() +
    theme(legend.title = element_blank())

  invisible(p)
}

my_plot <- plot.gtrends.silent(res)

my_plot +
  scale_x_date(date_breaks = "1 year", date_labels = "%Y") +
  theme(legend.position = "none")

The seasonal trend is due to when the NBA season starts and ends. The spike in 2014 is when the San Antonio Spurs were the NBA champions; the spike in 2013 is when they reached the NBA finals but lost.

Next, I’ll show how to plot interest per country; I’ll colour each city by the interest of a certain keyword. I will use the map_data function to obtain coordinates (and polygon drawing groups and orders) of cities in the world as a data frame. In the first example, I will use the keyword “wantok”, which means a close friend in Tok Pisin; this is the main language spoken in Papua New Guinea, the country I grew up in.

world <- map_data("world")

# change the region names to match the region names returned by Google Trends
world %>%
  mutate(region = replace(region, region=="USA", "United States")) %>%
  mutate(region = replace(region, region=="UK", "United Kingdom")) -> world

# perform search
res_world <- gtrends("wantok", time = "all")

# create data frame for plotting
res_world$interest_by_country %>%
  filter(location %in% world$region, hits > 0) %>%
  mutate(region = location, hits = as.numeric(hits)) %>%
  select(region, hits) -> my_df

ggplot() +
  geom_map(data = world,
           map = world,
           aes(x = long, y = lat, map_id = region),
           fill="#ffffff", color="#ffffff", size=0.15) +
  geom_map(data = my_df,
           map = world,
           aes(fill = hits, map_id = region),
           color="#ffffff", size=0.15) +
  scale_fill_continuous(low = 'grey', high = 'red') +
  theme(axis.ticks = element_blank(),
        axis.text = element_blank(),
        axis.title = element_blank())

As expected, the top interest in the keyword “wantok” is in Papua New Guinea.

We can also only focus on the United States. Here I’ll use the keyword swamp gravy, which was something I learned about while listening to the Planet Money podcast (highly recommended podcast!).

res <- gtrends("swamp gravy",
               geo = "US",
               time = "all")

state <- map_data("state")

res$interest_by_region %>%
  mutate(region = tolower(location)) %>%
  filter(region %in% state$region) %>%
  select(region, hits) -> my_df

ggplot() +
  geom_map(data = state,
           map = state,
           aes(x = long, y = lat, map_id = region),
           fill="#ffffff", color="#ffffff", size=0.15) +
  geom_map(data = my_df,
           map = state,
           aes(fill = hits, map_id = region),
           color="#ffffff", size=0.15) +
  scale_fill_continuous(low = 'grey', high = 'red') +
  theme(panel.background = element_blank(),
        axis.ticks = element_blank(),
        axis.text = element_blank(),
        axis.title = element_blank())

Swamp gravy is a play that has become the U.S. state of Georgia’s official folk-life play; I recommend the podcast on swamp gravy. The state of Georgia (and its neighbouring state Alabama) indeed has the highest interest in the keyword.

Since this is a bioinformatics blog, I’ll show results for “bioinformatics” using the same code as above but with the keyword replaced.

Maryland (location of the NIH) and Massachusetts (location of the Broad Institute) have the highest interest in bioinformatics.

Have a go with keywords of your choice! I hope you enjoyed this short post on plotting Google Trends results with R; have a Happy New Year!