Google trends datasets for product searches during COVID-19 crisis

covid-19-products-google-trends.R

library(data.table)
library(gtrendsR)
library(ggplot2)
library(ggthemes)

# Adopted from the product list by time.com:
# https://time.com/5810811/coronavirus-shopping-data/
products = c("milk", "meat", "alcohol",
             "cleaner", "beans", "toilet paper", 
             "hand sanitizer", "facial tissue", "rice",
             "first aid kit", "chickpeas", "paper towels",
             "black beans", "tuna", "soap", "soup", 
             "drinking water", "vitamins", "antibiotics", "chicken", "eggs",
             "popcorn", "bread", "cheese", "potato chips", "yogurt",
             "pastries", "ice cream", "apples",
             "chocolate", "bananas", "papayas", "celery", "pet toys",
             "vegetables", "sunscreen", "perfume")

gtrends_data = data.table(date = as.POSIXct(NA), geo = character(1), 
                          keyword = character(1), hits = integer(1))[!is.na(date)]
for(term in products) {
  tmp = gtrends(keyword = term, geo = c("US","CA","GB"), 
                time = "2019-12-28 2020-04-04", 
                gprop = "web", onlyInterest = TRUE)
  tmp$interest_over_time$hits = gsub("<1","0", tmp$interest_over_time$hits)
  gtrends_data = rbindlist(list(gtrends_data, tmp$interest_over_time[, c("date","geo","keyword","hits")]))
  Sys.sleep(1) # just in case
}

location = "data/product trends data/"
fwrite(gtrends_data, file = paste0(location, "products_demand.csv"))
fwrite(gtrends_data[date > "2020-03-26 19:00:00 CDT",], file = paste0(location, "products_demand_test.csv"))
fwrite(gtrends_data[date <= "2020-03-26 19:00:00 CDT",], file = paste0(location, "product_demand_train.csv"))

ggplot(gtrends_data, aes(date, hits, color=geo)) +
  geom_point(shape = 20) +
  facet_wrap(~keyword, ncol = 5) +
  scale_color_few(palette = "Medium", name = "Country") +
  theme_tufte(ticks = FALSE, base_family = "Palatino", base_size = 16) +
  theme(axis.text.y = element_blank(),
        axis.title.x = element_blank())