Jak Onet pracuje na Facebooku?

fb_page_posts <- get_fanpage_posts("Onet", 2000, token)

fb_page_posts %>%
   count(type) %>%
   mutate(p=100*n/sum(n)) %>%
   ungroup() %>%
   arrange(n) %>%
   mutate(type = factor(type, levels=type)) %>%
   ggplot() +
   geom_bar(aes(1, p, fill=type), stat="identity") +
   labs(fill="Typ postu") +
   theme_void() +
   coord_polar(theta="y")

fb_page_posts %>%
   count(hour, wday) %>%
   ungroup() %>%
   mutate(wday = factor(wday, rev(c("pn","wt","śr","cz","pt","sb","nd")))) %>%
   ggplot() +
   geom_tile(aes(hour, wday, fill=n)) +
   theme_minimal() +
   labs(x="Godzina", y="Dzień tygodnia") +
   scale_fill_gradient(low="#800026", high="#ffffcc") +
   theme(legend.position="none")

ggplot(fb_page_posts) +
   geom_smooth(aes(created_time, likes_count),
               color="red", se = FALSE) +
   geom_smooth(aes(created_time, comments_count),
               color="green", se = FALSE) +
   geom_smooth(aes(created_time, shares_count),
               color="blue", se = FALSE) +
   theme_minimal() +
   labs(title="Liczba interakcji na post", x="Czas",
        y="Liczba interakcji na post")

summary(fb_page_posts$likes_count)

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##     0.0    28.0    79.0   259.1   208.0 15800.0

summary(fb_page_posts$comments_count)

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    0.00    1.00    8.00   30.44   28.00  922.00

summary(fb_page_posts$shares_count)

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##     0.0     2.0    13.0    77.2    53.0  6387.0

get_comments <- function(post_id, token, n_comments=500, n_likes = 500) {

   # zabezpieczenie dla pętli wywołującej funkcję
   # poprawnie te warunki powinny być poza funkcją!
   if(n_comments == 0) n_comments <- 500
   if(n_likes == 0) n_likes <- 500
   
   coms <- getPost(post_id, token=token,
                   n.comments = n_comments,
                   n.likes = n_likes,
                   n = n_comments + n_likes)

   return(coms)
}


# pobierz komentarze i lajki do wszystkich postów
fb_comments <- data.frame()
fb_likes <- data.frame()
for(i in 266:nrow(fb_page_posts)) {
   post_data <- get_comments(fb_page_posts[i, "id"], token, 
                             fb_page_posts[i, "comments_count"],
                             fb_page_posts[i, "likes_count"])
   
   
   fb_comments_tmp <- post_data$comments
   # czy są jakieś komentarze?
   # jeśli tak - dodaj do listy wszystkich
   if(!is.null(fb_comments_tmp))
      if(nrow(fb_comments_tmp) > 0) {
         fb_comments_tmp$post_id <- fb_page_posts[i, "id"]
         fb_comments <- rbind(fb_comments, fb_comments_tmp)
      }
   

   fb_likes_tmp <- post_data$likes
   # czy są jakieś likes?
   # jeśli tak - dodaj do listy wszystkich
   if(!is.null(fb_likes_tmp))
      if(nrow(fb_likes_tmp) > 0) {
         fb_likes_tmp$post_id <- fb_page_posts[i, "id"]
         fb_likes <- rbind(fb_likes, fb_likes_tmp)
   }
}
# data w przyjaznym formacie
fb_comments$created_time = ymd_hms(fb_comments$created_time)

n_posts <- nrow(fb_page_posts)

fb_likes %>%
   group_by(from_id) %>%
   mutate(n=n()) %>%
   ungroup() %>%
   select(from_id, n) %>%
   unique() %>%
   mutate(p = 100*n/n_posts) %>%
   arrange(desc(n)) %>%
   head(10)

fb_comments %>%
   group_by(from_id) %>%
   mutate(n=n()) %>%
   ungroup() %>%
   select(from_id, n) %>%
   unique() %>%
   mutate(p = 100*n/n_posts) %>%
   arrange(desc(n)) %>%
   head(10)

fb_comments %>%
   filter(from_id =="1466984766960019") %>%
   count(post_id) %>%
   ungroup() %>%
   summary(n)

##    post_id                n        
##  Length:313         Min.   :1.000  
##  Class :character   1st Qu.:1.000  
##  Mode  :character   Median :1.000  
##                     Mean   :1.003  
##                     3rd Qu.:1.000  
##                     Max.   :2.000

fb_comments %>%
   count(from_id, post_id) %>%
   arrange(desc(n))

fb_comments %>%
   filter(post_id=="185265414715_10154394754944716",
          from_id=="1409834469287770") %>%
   select(created_time, message) %>%
   arrange(created_time)

# same komentarze i daty
com_times <- fb_comments %>%
   transmute(post_id=post_id, created_time_com=created_time)

post_times <- fb_page_posts %>%
   transmute(post_id=id, created_time_post=created_time)

# złączmy to po ID postu
times <- left_join(com_times, post_times, by="post_id") %>%
   mutate(diff = floor(as.numeric(created_time_com - created_time_post) / 60)) %>%
   filter(diff >= 0)

times %>% 
   filter(diff <= 24*60) %>%
   ggplot() +
   geom_histogram(aes(diff), bins = 24*60) +
   geom_vline(xintercept = 60*(1:24), color="red") +
   theme_minimal()

quantile(floor(times$diff/60), seq(0, 1, 0.1))

##   0%  10%  20%  30%  40%  50%  60%  70%  80%  90% 100% 
##    0    0    0    1    2    3    6    9   14   22  795

times %>% 
   filter(diff <= 60) %>%
   ggplot() +
   geom_histogram(aes(diff), bins = 60) +
   scale_x_continuous(breaks = seq(0, 60, 5))

fb_comments %>%
   mutate(hour = hour(created_time)) %>%
   count(hour) %>%
   ungroup() %>%
   ggplot() +
   geom_bar(aes(hour, n), stat="identity") +
   theme_minimal() +
   labs(x="Godzina", y="Łączna liczba komentarzy")

fb_comments %>%
   mutate(hour = hour(created_time),
          wday = wday(created_time)) %>%
   mutate(wday = factor(wday,
                        levels=c(2,3,4,5,6,7,1),
                        labels=c("pn","wt","śr","cz","pt","sb","nd"))) %>%
   count(wday, hour) %>%
   ungroup() %>%
   ggplot() +
   geom_tile(aes(hour, wday, fill=n)) +
   theme_minimal() +
   labs(x="Godzina", y="Dzień tygodnia") +
   scale_fill_gradient(low="#800026", high="#ffffcc") +
   theme(legend.position="none")

fb_comments %>%
   mutate(hour = hour(created_time),
          wday = wday(created_time)) %>%
   mutate(wday = factor(wday,
                        levels=c(2,3,4,5,6,7,1),
                        labels=c("pn","wt","śr","cz","pt","sb","nd"))) %>%
   filter(wday=="pt", hour==18) %>%
   count(post_id) %>%
   ungroup() %>%
   arrange(desc(n))

fb_page_posts[fb_page_posts$id=="185265414715_10154420101434716", "message"]

## [1] "Premier trafiła do szpitala"

library(tidytext)

words_comments <- fb_comments %>%
   transmute(message = tolower(iconv(message, "UTF-8", "LATIN2"))) %>%
   unnest_tokens(words, message) %>%
   count(words) %>%
   ungroup() %>%
   filter(nchar(words) > 2)
   
words_posts <- fb_page_posts %>%
   transmute(message = tolower(iconv(message, "UTF-8", "LATIN2"))) %>%
   unnest_tokens(words, message) %>%
   count(words) %>%
   ungroup() %>%
   filter(nchar(words) > 2)

pl_stop_words <- readLines("../textmining/polish_stopwords.txt", encoding = "UTF-8")
words_comments <- filter(words_comments, !words %in% pl_stop_words)
words_posts <- filter(words_posts, !words %in% pl_stop_words)

library(wordcloud)
library(RColorBrewer)
pal <- brewer.pal(12, "Paired")

wordcloud(words_posts$words, words_posts$n,
          scale=c(1.9,.9), colors = pal,
          max.words = 100)

wordcloud(words_comments$words, words_comments$n,
          scale=c(1.9,.9), colors = pal,
          max.words = 100)