Fred Truman
Design & Code Creative
Handling Twitter Data with R
September 26, 2011
I’m in a class where we are learning R. Here’s my second week’s assignment. – The number of users with more than 10000 followers. What percentage of the total number of users is this?
tonsoffollowers[tonsoffollowers$followers_count 10000,]
>>>
2322
1.3%- The user’s name and screen name with the most followers
max(tonsoffollowers$followers_count)
tonsoffollowers[tonsoffollowers$followers_count==2562105,]
>>>
Rev Run
RevRunWisdom
2562105 followers- The number of users who are following 0 people (the friends_count variable).
users[users$followers_count==0,]
>>
7110- The percentage of accounts that are protected.
users[users$protected=="true",]/nrow(users)
>>>
1.6% (28,176)- The number of people who have more followers than friends.
users[users$followers_count>users$friends_count,]
>>>
75258- The number of people who have exactly the same number of followers as friends.
6367- The user’s name and screen name who is listed the most times.
max(users$listed)
>>>
33636
which(users$listed==33636)
>>>
Miley Ray Cyrus
MileyCyrus
listed 336362. Because I can’t get away from those baby names, create a plot of the counts of the first letters of Twitter user names. Let’s make a crazy assumption that the average Twitter user’s age is 22, so compare this plot with the first letters of baby names from 2011 – 22 = 1989. Write a couple of sentences about the differences / similarities you see. HINT: You should use tolower() at some point, and you’ll want to strip out the Unicode characters that don’t correspond to a-z. R has a handy built-in vector called “letters” that you can use to easily strip out the first letters you have that are not in az.
gsub("[^a-zA-Z]", "", x)
first.letter <- substr(gsub("[^a-zA-Z]", "",tolower(users$name)),1,1)(I realize I didn't answer this here... lost my history and need to redo)
3. Let’s look at how the number of tweets you have and the number of followers you have relate. Plot the log10() of followers_count vs. log10() of statuses_count with more appropriate x and y axis labels and a nice title. Color the points based on whether the account is protected or not. Do you see any relationship between the followers count and statuses count? What can we learn about the nature of protected account from this graph?
users <- read.csv("users.csv", h=T, stringsAsFactors=FALSE)
users.protected <- users[users$protected=="true",]
users.notprotected <- users[users$protected=="false",]
#
par(bg = "gray15")
par(fg = "gray30")
par(col.lab = "gray30")
par(col.main = "gray50")
plot_colors <- c("#FF000005","#0000FF05")
legend_colors <- c("#FF000090","#0000FF90")
plot(log10(users.notprotected$followers_count), log10(users.notprotected$statuses_count), pch=16, bty="n", xlab="Number of Followers (log10)", ylab="Number of Tweets (log10)", main="Twitter Users", col=plot_colors[2])
points(log10(users.protected$followers_count), log10(users.protected$statuses_count), pch=16, col=plot_colors[1])
legend("topleft", c("Protected","Not Protected"), cex=0.6, bty="n", pch=16, col=legend_colors);Interesting resources I found while doing this: