Visualize your mentions over time

tweets.R

library(ggplot2)
library(lubridate)
library(dplyr)
library(reshape2)
myname <- "@vsbuffalo" # for removing later

d <- read.csv("tweets.csv", header=TRUE, stringsAsFactors=FALSE)

extractMentions <- function(x) {
    gsub("[^@]*(@[a-zA-Z0-9_]+).*", "\\1", x, perl=TRUE)
}

getMentions <- function(x) {
	words <- strsplit(x, " +")
	mentions <- lapply(words, function(w) {
		        extractMentions(w[grep("@", w)])
                    })
	all_users  <- sort(unique(unlist(mentions)))
	mentions <- lapply(mentions, function(m) factor(m, levels=all_users))

	tmp <- do.call(rbind, lapply(mentions, table))
	tmp
}

monyr <- function(x) {
	x <- as.POSIXlt(x)
	x$mday <- 1
	as.Date(x)
}

mentions <- getMentions(d$text)

csums_mentions <- colSums(mentions)

# tweek how many folks you see here:
# mentions_subset <- mentions[, csums_mentions > quantile(csums_mentions, probs=0.98)]
mentions_subset <- mentions[, csums_mentions > 50]

times <- parse_date_time(d$timestamp, "ymd_hms z*!")
dm <- data.frame(time=times, year=year(times), month=month(times))

dm <- cbind(dm, mentions_subset)
dmelt <- melt(dm, id.vars=c('time', 'year', 'month'))

xx <- dmelt %.% group_by(monyr=monyr(time), variable) %.% summarise(ntweets = sum(value))

p <- ggplot(xx[xx$variable != myname,])
p <- p + geom_line(aes(x=monyr, group=variable, color=variable, y=ntweets))
p <- p +  xlab("") + ylab("tweets per month") + scale_color_discrete("mentions")
p