Joshua Cook May 26, 2020
TidyTuesday link: https://github.com/rfordatascience/tidytuesday/blob/master/data/2020/2020-05-26/readme.md
knitr::opts_chunk$set(echo = TRUE, comment = "#>")
library(glue)
library(lubridate)
library(patchwork)
library(magrittr)
library(tidytext)
library(testthat)
library(tidyverse)
library(conflicted)
conflict_prefer("filter", "dplyr")
conflict_prefer("select", "dplyr")
theme_set(theme_minimal())# cocktails <- read_csv("https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2020/2020-05-26/cocktails.csv")
boston_cocktails <- read_csv("https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2020/2020-05-26/boston_cocktails.csv") %>%
janitor::clean_names()#> Parsed with column specification:
#> cols(
#> name = col_character(),
#> category = col_character(),
#> row_id = col_double(),
#> ingredient_number = col_double(),
#> ingredient = col_character(),
#> measure = col_character()
#> )
boston_cocktails %>%
count(ingredient) %>%
ggplot(aes(x = log2(n))) +
geom_density(size = 1, fill = "skyblue3", alpha = 0.2) +
scale_x_continuous(expand = expansion(mult = c(0, 0.02))) +
scale_y_continuous(expand = expansion(mult = c(0, 0.02))) +
labs(x = "number of times the ingredient is used (log2-transformed)",
y = "density",
title = "A long-tail distribution for the number of recipes an ingredient is in") 
boston_cocktails %>%
summarise(num_drinks = n_distinct(name),
num_categories = n_distinct(category),
num_ingredients = n_distinct(ingredient))#> # A tibble: 1 x 3
#> num_drinks num_categories num_ingredients
#> <int> <int> <int>
#> 1 989 11 569
boston_cocktails %>%
group_by(category) %>%
summarise(num_ingredients = n_distinct(ingredient),
num_drinks = n_distinct(name)) %>%
ungroup() %>%
arrange(num_ingredients)#> # A tibble: 11 x 3
#> category num_ingredients num_drinks
#> <chr> <int> <int>
#> 1 Non-alcoholic Drinks 3 1
#> 2 Rum 4 1
#> 3 Shooters 4 2
#> 4 Cordials and Liqueurs 15 8
#> 5 Gin 29 17
#> 6 Brandy 59 47
#> 7 Rum - Daiquiris 126 111
#> 8 Tequila 129 87
#> 9 Whiskies 142 110
#> 10 Vodka 179 138
#> 11 Cocktail Classics 228 467
boston_cocktails %>%
count(name, category, name = "num_ingredients") %>%
group_by(category) %>%
filter(n_distinct(name) > 20) %>%
ungroup() %>%
ggplot(aes(x = num_ingredients)) +
geom_bar(aes(fill = category, color = category),
size = 1, alpha = 0.7) +
scale_color_brewer(palette = "Set1") +
scale_fill_brewer(palette = "Set1") +
scale_x_continuous(expand = expansion(mult = c(0, 0.02)),
breaks = 1:6) +
scale_y_continuous(expand = expansion(mult = c(0, 0.02))) +
labs(x = "number of ingredients per drink",
y = "count",
title = "The distribution of the number of ingredients per drink") 
most_common_ingredients <- boston_cocktails %>%
select(category, ingredient) %>%
mutate(ingredient = str_to_lower(ingredient)) %>%
unnest_tokens(word, ingredient) %>%
anti_join(stop_words, by = "word") %>%
count(category, word) %>%
group_by(category) %>%
top_n(10, wt = n) %>%
ungroup()parse_ounces <- function(val) {
if (val == "For glass") { return(1) }
if (str_detect(val, "or")) {
val <- str_remove(val, "or.+$")
}
val_pieces <- str_remove_all(val, "oz|slices|tsp|dash") %>%
str_trim() %>%
str_split(" ") %>%
unlist() %>%
str_trim()
val_pieces <- val_pieces[str_length(val_pieces) > 0]
map_dbl(val_pieces, function(x){
if (str_detect(x, "/")) {
x_frac <- str_split(x, "/") %>% unlist() %>% as.numeric()
return(x_frac[[1]] / x_frac[[2]])
} else {
return(as.numeric(x))
}
}) %>%
sum()
}
test_that("The ounces are parsed correctly.", {
expect_equal(parse_ounces("1 oz"), 1)
expect_equal(parse_ounces("1/2 oz"), 0.5)
expect_equal(parse_ounces("1 1/2 oz"), 1.5)
expect_equal(parse_ounces("For glass"), 1)
expect_equal(parse_ounces("1/2 or 1"), 0.5)
expect_equal(parse_ounces("1 3/4 oz"), 1.75)
expect_equal(parse_ounces("1 dash"), 1)
})cocktails <- boston_cocktails %>%
group_by(category) %>%
filter(n_distinct(name) > 20) %>%
ungroup() %>%
mutate(ingredient = str_to_lower(ingredient)) %>%
unnest_tokens(word, ingredient) %>%
inner_join(most_common_ingredients,
by = c("category", "word")) %>%
group_by(name, category, word) %>%
slice(1) %>%
ungroup()
cocktails %<>%
mutate(amount = map_dbl(measure, parse_ounces))
stopifnot(!any(is.na(cocktails$amount)))cocktails_wide <- cocktails %>%
pivot_wider(c(name, category), names_from = word, values_from = amount) %>%
mutate_if(is.numeric, function(x) ifelse(is.na(x), 0, x))library(Rtsne)
pca_data <- cocktails_wide %>%
select(-c(name, category)) %>%
as.data.frame()
rownames(pca_data) <- cocktails_wide %>%
mutate(rowname = paste0(name, "___", category) %>% str_replace_all(" ", "_")) %>%
pull(rowname)
pca_data <- unique(pca_data)
drinks_pca <- prcomp(pca_data, center = TRUE, scale. = TRUE)
tsne_drinks <- Rtsne(drinks_pca$x[], perplexity = 5)
tsne_drinks_tib <- as_tibble(tsne_drinks$Y) %>%
set_names(c("z1", "z2")) %>%
mutate(rowname = rownames(pca_data)) %>%
separate(rowname, into = c("name", "category"), sep = "___") %>%
mutate(name = str_replace_all(name, "_", " "),
category = str_replace_all(category, "_", " "))#> Warning: The `x` argument of `as_tibble.matrix()` must have column names if `.name_repair` is omitted as of tibble 2.0.0.
#> Using compatibility `.name_repair`.
#> This warning is displayed once every 8 hours.
#> Call `lifecycle::last_warnings()` to see where this warning was generated.
tsne_drinks_summary <- tsne_drinks_tib %>%
group_by(category) %>%
summarise(avg_z1 = mean(z1), avg_z2 = mean(z2)) %>%
ungroup()
tsne_drinks_tib %>%
ggplot(aes(z1, z2)) +
geom_point(aes(color = category)) +
geom_label(aes(label = category, x = avg_z1, y = avg_z2),
data = tsne_drinks_summary,
label.size = 0,
label.padding = unit(1, "mm"),
fill = "white",
fontface = "bold",
alpha = 0.8) +
scale_color_brewer(palette = "Set2") +
theme(
plot.title = element_text(hjust = 0.5),
legend.position = "none"
) +
labs(x = "t-SNE dim. 1",
y = "t-SNE dim. 2",
title = "PCA and t-SNE of drinks by ingredients")