Number of graphs / frequency of tweets per word per month
I hooked R to Twitter and scrape using a function searchTwitter
in R and scrape the resulting data for punctuation, lowercase letters, etc. Now I am trying to do the following:
- count the number of tweets with the word "auction" that were tweeted per month from January 2015 to the end of July 2015.
- Set up the graph on a simple histogram (
x-axis - month
;y-axis - number of tweets
).
I would like to reuse this for retweets, mentions, replies, and favorites.
This is what I have tried so far:
#load the packages into R
>library(twitteR)
>library(plyr)
>library(ggplot2)
# Register an application (API) at https://apps.twitter.com/
# Look up the API key and create a token – you need for both the key and the secret
# Assign the keys to variables and use the authorization
api_key <- "your API key from twitter"
api_secret <- "your Secret key from twitter"
access_token <- "you Access Token from twitter"
access_token_secret <- "you Access Token Secret key from twitter"
setup_twitter_oauth(api_key,api_secret,access_token,access_token_secret)
1 "Using Direct Authentication" Use a local file to cache OAuth access credentials between R sessions?
1: Yes
2: No
# Type 1 and press Enter. Choice: 1
auctiontweets <- searchTwitter("auction", since = "2015-01-01", until = "2015-08-03", n=1000)
However, I am unable to create the dataframe, getting the following error:
tweet.dataframe <- data.frame(searchTwitter("action", since = "2015-01-01", until = "2015-08-03", n=3000))
Error in as.data.frame.default (x [[i]], optional = TRUE):
cannot force class "structure" ("status", package = "twitteR") "to data.frame
I found some code on how to set up users by the hour; but couldn't change it to work for tweets with a specific word (ie "auction") per month:
yultweets <- searchTwitter("#accessyul", n=1500)
y <- twListToDF(yultweets)
y$created <- as.POSIXct(format(y$created, tz="America/Montreal"))
yply <- ddply(y, .var = "screenName", .fun = function(x) {return(subset(x,
created %in% min(created), select = c(screenName,created)))})
yplytime <- arrange(yply,-desc(created))
y$screenName=factor(y$screenName, levels = yplytime$screenName)
ggplot(y) + geom_point(aes(x=created,y=screenName)) + ylab("Twitter username") + xlab("Time")
The source can be found here .
source to share
Since you haven't provided even a small piece of your data that we can handle, my answer may be superficial.
library(stringi); library(dplyr); library(SciencesPo)
df <- data.frame(tweets = c("blah, blah, Blah, auction","blah, auction", "blah, blah", "this auction, blah", "today"), date=c('2015-07-01','2015-06-01','2015-05-01','2015-07-31','2015-05-01'))
> df
tweets date
1 blah, blah, Blah, auction 2015-07-01
2 blah, auction 2015-06-01
3 blah, blah 2015-05-01
4 this auction, blah 2015-07-31
5 today 2015-05-01
filter = "auction"
> df$n <- vapply(df$tweets, function(x) sum(stri_count_fixed(x, filter)), 1L)
> df
tweets date n
1 blah, blah, Blah, auction 2015-07-01 1
2 blah, auction 2015-06-01 1
3 blah, blah 2015-05-01 0
4 this auction, blah 2015-07-31 1
5 today 2015-05-01 0
Then you only need to summarize:
df %>% group_by(month=format(as.Date(date),format="%m")) %>% summarize(freq=sum(n))
%>%ungroup() -> df2
> df2
Source: local data frame [3 x 2]
month freq
1 05 0
2 06 1
3 07 2
>
Voila! Bonus, write it down likeggplot(df2, aes(x=month, y=freq)) + geom_line() + theme_pub()
source to share