Create Articles vs Reactions Joyplot Animation

create-joyplot.R

# Creating joyplot at: https://github.com/tjpalanca/facebook-news-analysis/raw/master/figs/09-reactions-vs-articles.gif
# Full rmarkdown file at: https://github.com/tjpalanca/facebook-news-analysis/blob/master/src/02-fb-topic-modeling.Rmd

# Prepare dataset for joy plot
posts_classification.dt %>% 
  inner_join(
    posts.dt %>% 
      select(
        document = post_id, 
        news_page = page_name, 
        attachment_title, 
        post_timestamp_utc
      ),
    by = "document"
  ) %>% 
  inner_join(topics_mapping.dt, by = "topic") %>% 
  filter(
    !topic_name %in% 
      c("Filipino News", "Filipino Showbiz", 
        "Filipino Sports", "Breaking News")
  ) %>%
  inner_join(
    reactions.dt %>% 
      group_by(document = object_id) %>%
      summarise(reactions = sum(count)) %>%
      ungroup(),
    by = "document"
  ) %>%
  group_by(
    topic_name, 
    # 8 hour time difference (Manila)
    week = floor_date(post_timestamp_utc + hours(8), "week")
  ) %>% 
  summarise(
    num_articles  = sum(allocation),
    num_reactions = sum(reactions * allocation)
  ) %>%
  mutate(peak = week[num_articles == max(num_articles)]) %>%
  ungroup() %>%
  arrange(peak, week) %>%
  mutate(
    topic_name = factor(topic_name, levels = rev(unique(topic_name))),
    # ID needed for linking back information after tweening frames
    id = as.character(row_number()),
    # Maximum height for ridgeline is 9
    articles_height = num_articles/max(num_articles) * 9,
    reactions_height = num_reactions/max(num_reactions) * 9
  ) ->
  posts_time_trend.dt

# Prepare animated dataset
bind_rows(
  # Articles
  posts_time_trend.dt %>% 
    transmute(id, time = 0,  ht = articles_height , op = 0),
  # Reactions
  posts_time_trend.dt %>% 
    transmute(id, time = 10, ht = reactions_height, op = 1),
  # Return to articles
  posts_time_trend.dt %>% 
    transmute(id, time = 20, ht = articles_height , op = 0),
) %>%
  # Add cubic easing
  mutate(ease = "cubic-in-out") %>%
  tween_elements("time", "id", "ease") %>%
  # Add back the topic name and week
  inner_join(
    posts_time_trend.dt %>% select(topic_name, week, id), 
    by = c(".group" = "id")
  ) ->
  posts_time_trend.adt

# Create labels above the chart (what news outlets post vs what we react to)
posts_time_trend.adt %>%
  group_by(.frame) %>%
  summarise(
    opacity    = mean(op),
    range_week = max(week) - min(week),
    min_week   = min(week),
    num_topics = n_distinct(topic_name)
  ) %>% 
  ungroup() -> 
  posts_time_trend_labels.adt


# Create animated pot
posts_time_trend.adt %>%
  ggplot(aes(x = week, y = topic_name, height = ht, 
             fill = topic_name, frame = .frame)) +
  geom_ridgeline(color = NA, alpha = 0.5) +
  geom_text(
    data = posts_time_trend_labels.adt,
    aes(x = min_week + 0.25 * range_week, 
        y = num_topics + 3.5, 
        alpha = 1 - opacity, 
        frame = .frame),
    label = "What news outlets publish", color = "forestgreen",
    family = "Roboto", fontface = "bold", size = 9,
    inherit.aes = FALSE
  ) +
  geom_text(
    data = posts_time_trend_labels.adt,
    aes(x = min_week + 0.75 * range_week, 
        y = num_topics + 3.5, 
        alpha = opacity, 
        frame = .frame),
    label = "What we react to", color = "firebrick",
    family = "Roboto", fontface = "bold", size = 9,
    inherit.aes = FALSE
  ) +
  geom_text(
    data = posts_time_trend_labels.adt,
    aes(x = min_week + 0.55 * range_week, 
        y = num_topics + 3.5, 
        frame = .frame),
    label = "VS",
    family = "Roboto", fontface = "bold", size = 7,
    inherit.aes = FALSE
  ) +
  scale_x_datetime(expand = c(0, 0)) +
  theme(
    legend.position = "none",
    panel.border    = element_blank(),
    axis.title      = element_blank(),
    panel.grid.major.x = element_blank(),
    panel.grid.minor.x = element_blank(),
    plot.caption      = element_text(size = 9)
  ) + 
  labs(
    caption = "
DATA SOURCE: Facebook Graph API
CHART NOTE: Topics were derived through Latent Dirichlet Allocation (LDA) with 40 topics for Facebook news articles in 2016 for the Philippines
CHART BY: TROY JAMES PALANCA | TJPALANCA.COM
    "
  ) ->
  posts_time_trend.agg

gganimate(
  p = posts_time_trend.agg,
  filename = "../figs/09-reactions-vs-articles.gif",
  title_frame = FALSE,
  interval = 1/10,
  ani.width = 900,
  ani.height = 600
)