Top Tweet by User

First, we want to look at who is the top user by looking at their count of retweet.

tweet %>% arrange(desc(retweet_count)) %>% select(screen_name, name, location, 
    text, retweet_count, created_at) %>% mutate(date = date(created_at)) %>% 
    slice(1:10) %>% ggplot(aes(reorder(name, retweet_count), retweet_count, 
    fill = retweet_count)) + geom_col() + coord_flip() + theme_minimal() + scale_fill_viridis_c(guide = F) + 
    labs(title = "Top Tweet related to Grab and Gojek Indonesia", subtitle = "Excluding any retweet text", 
        y = "Retweet Count", x = "Name")

Let’s see what they tweeted about.

tweet %>% arrange(desc(retweet_count)) %>% select(screen_name, name, location, 
    text, retweet_count, created_at) %>% mutate(date = date(created_at)) %>% 
    slice(1:10) %>% mutate(text = text %>% replace_url() %>% replace_html() %>% 
    replace_tag() %>% replace_contraction() %>% replace_word_elongation() %>% 
    str_squish()) %>% select(screen_name, retweet_count, text)

Where most of the tweets came from

Next, we want to see where most of the tweets came from.

tweet %>% mutate(location = case_when(str_detect(location, "Jakarta") ~ "Jakarta", 
    str_detect(location, "Bandung") ~ "Bandung", str_detect(location, "62") ~ 
        "Indonesia", str_detect(location, "Yogyakarta") ~ "Yogyakarta", TRUE ~ 
        location), location = str_replace(location, "indonesia", "Indonesia"), 
    location = if_else(location == "", "Unknown", location)) %>% group_by(location) %>% 
    summarise(total = n(), retweet = sum(retweet_count)) %>% arrange(desc(total)) %>% 
    slice(1:15) %>% ggplot(aes(reorder(location, total), total, fill = retweet)) + 
    geom_col() + coord_flip() + theme_minimal() + scale_fill_viridis_c(option = "B") + 
    labs(title = "Where Most of the Tweets Came From", x = "Location", y = "Total Tweets") + 
    theme(legend.position = "bottom")

The location is retrieved via the profiles in the bio of each twitter accounts. Turns out most of people didn’t put their location in the bio, while those who put their location are mostly come from various area of Jakarta and Bandung or only wrote Indonesia.

Retweets

Here, we will build a network of retweets. This graph linking users to the users they retweet to fundamentally visualise how information spreads throughout Twitter.

tweet_re <- search_tweets(q = "@GrabID OR @gojekindonesia", n = 10000, include_rts = T, 
    token = token)

net <- tweet_re %>% gt_edges(screen_name, retweet_screen_name) %>% gt_nodes() %>% 
    gt_collect()

net$nodes <- net$nodes %>% mutate(color = case_when(nodes == "grabid" ~ "#fca510", 
    nodes == "gojekindonesia" ~ "#d91a1a", TRUE ~ "#1dcaff"))

net$edges <- net$edges %>% mutate(color = case_when(target == "grabid" ~ "#fff2a1", 
    target == "gojekindonesia" ~ "#f75959", TRUE ~ "#1dcaff"))

c(edges, nodes) %<-% net
nodes <- nodes2sg(nodes)
edges <- edges2sg(edges)

nodes %>% arrange(desc(size))

sigmajs() %>% sg_nodes(nodes, id, label, size, color) %>% sg_edges(edges, id, 
    source, target, color) %>% sg_layout(layout = igraph::layout_components) %>% 
    sg_settings(minNodeSize = 1, maxNodeSize = 5)

Red point indicate gojekindonesia and the red line indicate which account retweeted from gojekindonesia, while orange point indicate GrabID.

There is a lot of people who retweeted GrabID, while for the gojekindonesia, not so much. More people talk about GrabID than gojek. Let’s see what people are talking about GrabID.

tweet_re %>% filter(str_detect(text, pattern = "GrabID")) %>% mutate(retweet_text = retweet_text %>% 
    replace_url() %>% replace_html() %>% replace_tag() %>% replace_contraction() %>% 
    replace_word_elongation() %>% str_squish()) %>% group_by(retweet_name, retweet_text) %>% 
    summarise(retweet = sum(retweet_count)) %>% arrange(desc(retweet)) %>% select(retweet_name, 
    retweet, retweet_text) %>% na.omit()

The most retweeted text is about a driver of GrabID who lost his motorcycle and the user Pernah Punya Nama is asking people to help him. The second one is about a scandal related to the user Iman Sjafei, who tweeted that GrabID stole his marketing campaign ideas. The third one is about a theft done by a driver from gojekindonesia to one of the GrabID customer, while the fourth one is about a touching story about a customer with her GrabID driver. Based on these top 4 tweets, there is a mixed of a bad and good news related to GrabId.

How about tweets related to gojekindonesia?

tweet_re %>% filter(str_detect(text, pattern = "gojekindonesia")) %>% mutate(retweet_text = retweet_text %>% 
    replace_url() %>% replace_html() %>% replace_tag() %>% replace_contraction() %>% 
    replace_word_elongation() %>% str_squish()) %>% group_by(retweet_name, retweet_text) %>% 
    summarise(retweet = sum(retweet_count)) %>% arrange(desc(retweet)) %>% select(retweet_name, 
    retweet, retweet_text) %>% na.omit()

The top 2 tweets is about a touching story related to the gojekindonesia drivers, while the next 2 tweets is about a bad experience about the gojekindonesia. Both company need to be more concerned about their drivers since social media can really hurt the reputation of the company, not just the driver partner.

Further analysis can be done to see if the user or most of the retweet is bot.

Hashtags

Here, we will build a network of hastags. Let’s see what hastags is used simultaneously.

net <- tweet %>% gt_edges(screen_name, hashtags) %>% gt_nodes() %>% gt_collect()

c(edges, nodes) %<-% net

nodes <- nodes2sg(nodes)
edges <- edges2sg(edges)

nodes$color <- ifelse(nodes$type == "user", "#0084b4", "orange")

sigmajs() %>% sg_nodes(nodes, id, size, color, label) %>% sg_edges(edges, id, 
    source, target) %>% sg_layout(layout = igraph::layout_components) %>% sg_settings(edgeColor = "default", 
    defaultEdgeColor = "#d3d3d3") %>% sg_neighbours()

Orange point indicate the hastags while the blue one indicate who tweeted the hastags.

The biggest hastags is #Selalubisa, which is used alongside #siapgrab and #aplikasiuntuksemua. These hastags is invented to promote the newest GrabID marketing campaign. Meanhwile, gojekindoesia didn’t have a single big hastags. The biggest one is only the #gojek and #pastiadajalan. Based on this analysis, people love to share their stories with GrabID compared to gojekindonesia. This may be inline with the research done by Grab, which stated that they has 70% of market share in four wheels service and 60% market share in two wheels service overall, while Gojek only won in the food delivery service (GoFood) with 72% of this market share.

Conclusions

To summarise the analysis, here is our conclusions so far:

• People love to share their experience with the driver partner of the ride hailing services, GrabID and gojekindoensia.
• The top retweeted text mostly consists of a touching stories or a scandal, warning or complaints
• GrabID gained more interaction from people compared to gojekindonesia.