Breizh Data Day Tweets

breizh-data-tweets.R

library(tidyverse)
library(rtweet)
library(proustr)
library(tidytext)

bdd <- search_tweets("#BreizhDataDay", n = 3000)

bdd_day <- bdd %>%
  filter(created_at < "2017-10-20 00:00:00", created_at > "2017-10-18 23:59:59")

# Tweet par heure

random_viridis <- function(x){
  sample(viridisLite::viridis(100), x)
}

ggplot(bdd_day) +
  aes(created_at) +
  geom_histogram(bins = 50, fill = random_viridis(1)) +
  labs(title = "#BreizhDataDay — Tweets dans la journée",
       x = "heure",
       y = "volume", 
       caption = "@breizhdataday") +
  theme_minimal()

# Tweets et retweets pendant la journée

ggplot(bdd_day) +
  aes(created_at, fill = is_retweet) +
  geom_density() +
  scale_fill_viridis_d() +
  labs(title = "Tweets et retweets pendant la journée",
       x = "",
       y = "Densité de tweets",
       caption = "@breizhdataclub") +
  facet_grid(is_retweet ~ .) +
  theme_minimal()

# Comptes les plus actifs

bdd_day %>%
  count(screen_name) %>%
  top_n(10) %>%
  ggplot() +
  aes(reorder(screen_name, n), n) +
  geom_col(fill = random_viridis(1)) +
  labs(title = "Comptes les plus actifs pendant le #BreizhDataDay",
       x = "",
       y = "Nombre de tweets",
       caption = "@breizhdataclub") +
  coord_flip() +
  theme_minimal()

# Comptes les plus cités

df <- unnest_tokens(bdd_day, comptes, mentions_screen_name) %>%
  count(comptes) %>%
  na.omit() %>%
  filter(n >= 5) %>%
  arrange(desc(n))

ggplot(df) +
  aes(reorder(comptes, n), n) +
  geom_col(fill = random_viridis(1)) +
  labs(title = "Comptes les plus cités pendant le #BreizhDataDay",
       subtitle = "5 fois ou plus",
       x = "",
       y = "volume",
       caption = "@breizhdataclub") +
  coord_flip() +
  theme_minimal()


# Hashtags les plus utilisés (hormis BreizhDataDay)

df <- unnest_tokens(bdd_day, hashtags, hashtags) %>%
  count(hashtags) %>%
  filter(n >= 5, hashtags != "breizhdataday") %>%
  arrange(desc(n))

ggplot(df) +
  aes(reorder(hashtags, n), n) +
  geom_col(fill = random_viridis(1)) +
  labs(title = "Hashtags accompagnant le plus #BreizhDataDay",
       x = "",
       y = "volume",
       caption = "@breizhdataclub") +
  coord_flip() +
  theme_minimal()

# Sentiment analysis 

sentiments <- proust_sentiments()

df <- unnest_tokens(bdd_day, word, text) %>%
  left_join(sentiments) %>%
  select(word, polarity, created_at) %>%
  na.omit()
ggplot(df) +
  aes(created_at, fill = polarity) +
  geom_histogram(bins = 50) + 
  facet_grid(polarity ~ .) + 
  scale_fill_manual(values = random_viridis(2)) + 
  labs(title = "Sentiment analysis pendant le #BreizhDataDay",
       x = "",
       y = "volume de tweets",
       caption = "@breizhdataclub") +
  theme_minimal()