R/track_twitter_hashtag.R
In AyrtonRua/group1_project: Package touristR (providing tourist friendly interactive map)
#' @title track_twitter_hashtag
#'
#' @description Provide a scatterplot displaying hashtag counts overtime
#' based on data obtained from Twitter.
#'
#' @param keyword A character \code{vector} of maximum length 4
#' (keywords requested)
#' @param type A character \code{vector} of length 1 (type of keyword, can
#' be either a city/place/monument name in English, to be referred to as
#' "place", or if another type of keyword, please specify by "other").
#' \cr
#' Note: all keywords should be of the same types (e.g. all places' names).
#' @param number A positive \code{integer} number (timeframe parameter)
#' @param sincetype A character \code{vector} of length 1 (timeframe parameter),
#' which can be (uniquely) on of the following:
#' either "days", "weeks", "months" or "years"
#' @param provideN A positive \code{integer} number
#' (Number ot tweets to fetch for each keyword)
#'
#' @return A timeseries scatterplot with the hashtag count on the y axis,
#' and on the x axis the timeframe.
#' @return A dataframe containing all the results fetched from Twitter
#' (e.g. tweets) based on package \code{\link{twitteR}}
#'
#' @author Ayrton Rua: \email{ayrton.gomesmartinsrua@unil.ch}
#' @author Maurizio Griffo: \email{maurizio.griffo@unil.ch}
#' @author Ali Karray: \email{mohamedali.karray@unil.ch}
#' @author Mohit Mehrotra: \email{mohit.mehrotra@unil.ch}
#' @author Youness Zarhloul: \email{youness.zarhloul@unil.ch}
#'
#' @references Orso, S., Molinari, R., Lee,
#' J., Guerrier, S., & Beckman, M. (2018).
#' An Introduction to Statistical Programming Methods with R.
#' Retrieved from \url{https://smac-group.github.io/ds/}
#'
#' @seealso \code{\link{touristR}} & \code{\link{run_shiny}}
#'
#' @examples \dontrun{
#' track_keyword(
#' keyword = c("eiffel tower", "san francisco", "london"),
#' type="place",
#' number = 2,
#' sincetype = "weeks",
#' provideN = 100
#' )
#' }
#'
#' @details The function automatically obtains the latitude and longitude of
#' each requested place, then queries the Twitter API to obtain the number of
#' tweets with the requested # keywords for the desired timeframe, and returns
#' at the end a scatterplot based on the
#' aesthetics of \code{\link{ggplot2}},
#' allowing thus to analyze which places are the most popular
#' over a given time period. \cr
#' It should be noted in that sense that if the time frame selected by the user
#' is less than or equal to 2 weeks, the plot returned groups data by hour,
#' whereas if the user requests longer timeframes the result will be grouped by
#' days for better representation of the changes over time
#' (short/long term). \cr \cr
#' Lastly, we advise the user to keep the parameter \code{provideN} low
#' to avoid long delay responses from Twitter API (e.g. setting provideN=100
#' would be preferable than 1000).
#'
#' @importFrom magrittr %>%
#'
#' @export
track_twitter_hashtag <-
function(keyword, type, number, sincetype, provideN) {
#making sure all inputs are correct
stopifnot({
is.vector(keyword, mode = "character")
length(keyword) <= 4
is.vector(type, mode = "character")
length(type) == 1
is.integer(number)
number > 0
is.character(sincetype)
length(sincetype) == 1
sincetype == "days" |
sincetype == "weeks" | sincetype == "months" |
sincetype == "years"
is.integer(provideN)
provideN > 0
})
#accessing Twitter via API
consumer_key <- 'ugjqg8RGNvuTAL1fEiNtw'
consumer_secret <- '4WjuEbP6QLUN2DwDzyTwmdMES6fgnOsS65fWxpT8I'
access_token <-
'76887198-JA3xCVO1vvQMqMDiIobWKKGQxYKSB0CV2lI2PZ7GL'
access_secret <-
'IvzlVOC8KkIaMR5s5K4u2IXbxKQv7EcUSvy2bnaru8gKz'
#
twitteR::setup_twitter_oauth(consumer_key,
consumer_secret,
access_token,
access_secret)
#obtain the latitude and longitude of the keyword (city) input
#and save the result as a dataframe
#obtaining latitude and longitude for each place
df <- data.frame(city = NA,
long = NA,
lat = NA)
if (type == "place") {
for (i in 1:length(keyword)) {
df[i, 1] <- keyword[i]
#obtain latitude and longitude
df[i, 2:3] <- ggmap::geocode(location =
keyword[i] ,
source = "dsk",
messaging = FALSE)
}
} else if (type == "other") {
for (i in 1:length(keyword)) {
df[i, 1] <- keyword[i]
}
}
#vector to save the result (since when date)
sinceInput <- vector(mode = "character", length = 1L)
#adapting the parametersinceInput based on the requested timeframe
if (sincetype == "days") {
sinceInput[1] <-
(lubridate::today(tzone = Sys.timezone()) -
lubridate::days(number)) %>%
lubridate::ymd() %>% as.character()
} else if (sincetype == "weeks") {
sinceInput[1] <-
(lubridate::today(tzone = Sys.timezone()) -
lubridate::weeks(number)) %>% lubridate::ymd() %>% as.character()
} else if (sincetype == "months") {
#time difference in months (respecting e.g. months with 30 and 31 days)
sinceInput[1] <-
(lubridate::today(tzone = Sys.timezone()) -
lubridate::period(1, units = "month")) %>%
lubridate::ymd() %>% as.character()
} else if (sincetype == "years") {
sinceInput[1] <-
(lubridate::today(tzone = Sys.timezone()) -
lubridate::years(number)) %>%
lubridate::ymd() %>% as.character()
}
#saving results of for loop bellow
keyword_input <-
vector(mode = "character", length = length(keyword))
input <-
vector(mode = "character", length = length(keyword))
results_twitter = list() # saving result in a list
#querying Twitter including the geocode
#in order to make sure that the tweets are
#indeed related to the requested place
if (type == "place") {
for (i in 1:length(keyword)) {
#run query for that keyword (format is #keyword without space)
#first we remove any white space to obtain a one word input
keyword_input[i] <-
gsub(keyword[i], pattern = " ", replacement = "")
#then we obtain the #oneword format to be used as input for
#the Twitter query
input[i] <- paste("#", keyword_input[i], sep = "")
#query Twitter using the keyword input
results_twitter[[i]] <- twitteR::searchTwitter(
input[i] ,
n = provideN,
resultType = "mixed",
since = sinceInput[1] ,
until = lubridate::today(tzone = Sys.timezone()) %>%
lubridate::ymd() %>% as.character() ,
#specifying the geocode to be sure we only obtain e.g. indeed
#the results published from Paris for search query Paris (e.g.
#tweet should be made within a radius of 80 miles from Paris maximum)
geocode = paste(df[i, ]$lat, df[i, ]$long, "80mi", sep = ",")
)
} #querying twitter without geocode for type =="other"
} else if (type == "other") {
#query Twitter API
for (i in 1:length(keyword)) {
#run query for that keyword (format is #... without space)
#first we remove any white space to obtain a one word input
keyword_input[i] <-
gsub(keyword[i], pattern = " ", replacement = "")
#then we obtain the #oneword format to be used as input for
#the Twitter query
input[i] <- paste("#", keyword_input[i], sep = "")
#querying Twitter using the keyword input
results_twitter[[i]] <- twitteR::searchTwitter(
input[i] ,
n = provideN,
resultType = "mixed",
since = sinceInput[1] ,
until = lubridate::today(tzone = Sys.timezone()) %>%
lubridate::ymd() %>% as.character()
)
}
}
#obtaining the dataframe with the results fetched from Twitter
#obtaining the results based on the number of keywords requested
#between 1 and 4
if (length(keyword) == 4) {
results_twitter1 <-
results_twitter[[1]] %>% unlist() %>% twitteR::twListToDF() %>%
tibble::as.tibble() %>%
dplyr::mutate(city = keyword[1])
results_twitter2 <-
results_twitter[[2]] %>% unlist() %>% twitteR::twListToDF() %>%
tibble::as.tibble() %>%
dplyr::mutate(city = keyword[2])
results_twitter3 <-
results_twitter[[3]] %>% unlist() %>% twitteR::twListToDF() %>%
tibble::as.tibble() %>%
dplyr::mutate(city = keyword[3])
results_twitter4 <-
results_twitter[[4]] %>% unlist() %>% twitteR::twListToDF() %>%
tibble::as.tibble() %>%
dplyr::mutate(city = keyword[4])
finaldata <-
rbind(results_twitter1,
results_twitter2,
results_twitter3,
results_twitter4)
} else if (length(keyword) == 3) {
results_twitter1 <-
results_twitter[[1]] %>% unlist() %>% twitteR::twListToDF() %>%
tibble::as.tibble() %>%
dplyr::mutate(city = keyword[1])
results_twitter2 <-
results_twitter[[2]] %>% unlist() %>% twitteR::twListToDF() %>%
tibble::as.tibble() %>%
dplyr::mutate(city = keyword[2])
results_twitter3 <-
results_twitter[[3]] %>% unlist() %>% twitteR::twListToDF() %>%
tibble::as.tibble() %>%
dplyr::mutate(city = keyword[3])
finaldata <-
rbind(results_twitter1, results_twitter2, results_twitter3)
} else if (length(keyword) == 2) {
results_twitter1 <-
results_twitter[[1]] %>% unlist() %>%
twitteR::twListToDF() %>% tibble::as.tibble() %>%
dplyr::mutate(city = keyword[1])
results_twitter2 <-
results_twitter[[2]] %>% unlist() %>%
twitteR::twListToDF() %>% tibble::as.tibble() %>%
dplyr::mutate(city = keyword[2])
finaldata <- rbind(results_twitter1, results_twitter2)
} else if (length(keyword) == 1) {
results_twitter1 <-
results_twitter[[1]] %>% unlist() %>%
twitteR::twListToDF() %>% tibble::as.tibble() %>%
dplyr::mutate(city = keyword[1])
finaldata <- results_twitter1
}
#creating the plot of keyword count per day over time
#(measure of popularity)
if (sincetype == "days" && number <= 14) {
#correct date format and aggregate the results by hour which
#can be useful e.g. for marketing campaigns to study engagement
#levels for a certain keyword (e.g. a brand name) thus identifying
#the best time slots to advertise (i.e. when users are interacting
#the most with the brand).
finaldata$created <-
lubridate::ymd_hms(finaldata$created,
tz = Sys.timezone(),
quiet = TRUE) %>%
lubridate::round_date(unit = "hour")
finaldata$created_correct <-
as.POSIXct(format(finaldata$created, "%Y-%m-%d %H:%M:%S"),
format = "%Y-%m-%d %H")
#generating the plot
print(
finaldata %>%
dplyr::group_by(city, created_correct) %>% dplyr::count() %>%
dplyr::arrange() %>%
dplyr::rename(count = n) %>%
ggplot2::ggplot() +
ggplot2::geom_point(
mapping = ggplot2::aes(x = created_correct, count , color = city),
size = 3,
alpha = 0.8
) +
ggplot2::geom_line(ggplot2::aes(
x = created_correct, count , color = city
),
size = 0.5) +
ggplot2::scale_y_continuous(
breaks = scales::pretty_breaks(8),
labels = scales::number_format(accuracy = 1)
) +
ggplot2::labs(
title = "Twitter keyword popularity tracking",
caption = paste(
"Note: Hourly data (UTC time) fetched from",
format(as.Date(sinceInput[1]), "%A, %d-%b. %Y") ,
"until",
format(as.Date(Sys.Date()), "%A, %d-%b. %Y"),
".",
sep = " "
),
x = "Days",
y = "Number of tweets posted"
) +
ggplot2::theme_minimal() +
ggplot2::theme(
axis.title = ggplot2::element_text(size = 12, face = "bold"),
plot.caption = ggplot2::element_text(
size = 10,
face = "italic",
hjust = 0
),
plot.title = ggplot2::element_text(size = 14, face = "bold"),
axis.text.x = ggplot2::element_text(
size = 8,
angle = 25,
vjust = 1.0,
hjust = 1.0
),
axis.text.y = ggplot2::element_text(size = 8)
) +
ggplot2::theme(plot.margin = ggplot2::unit(c(2, 10, 2, 2), "mm")) +
viridis::scale_color_viridis(
discrete = TRUE ,
direction = -1,
alpha = 0.6
) +
ggplot2::scale_x_datetime(
labels = scales::date_format("%b %d - %H:%M"),
breaks = scales::date_breaks("5 hour")
) +
ggrepel::geom_text_repel(
data = finaldata %>%
dplyr::group_by(city, created_correct) %>%
dplyr::count() %>% dplyr::arrange(desc(n)) %>%
dplyr::filter(created_correct == lubridate::today()) %>%
dplyr::rename(count = n) %>%
dplyr::top_n(length(keyword)) ,
ggplot2::aes(x = created_correct,
count , label = factor(city))
,
fontface = 'bold',
color = 'black',
box.padding = 1,
point.padding = 1,
segment.color = 'black' ,
size = 4
) + ggplot2::guides(color = ggplot2::guide_legend(title = "Place"))
)
} else if (sincetype == "weeks" && number < 3) {
#correct date format and aggregate by hour
finaldata$created <-
lubridate::ymd_hms(finaldata$created,
tz = Sys.timezone(),
quiet = TRUE) %>%
lubridate::round_date(unit = "hour")
finaldata$created_correct <-
as.POSIXct(format(finaldata$created, "%Y-%m-%d %H:%M:%S"),
format = "%Y-%m-%d %H")
#plot
print(
finaldata %>%
dplyr::group_by(city, created_correct) %>%
dplyr::count() %>% dplyr::arrange() %>%
dplyr::rename(count = n) %>%
ggplot2::ggplot() +
ggplot2::geom_point(
mapping = ggplot2::aes(x = created_correct, count , color = city),
size = 3,
alpha = 0.8
) +
ggplot2::geom_line(ggplot2::aes(
x = created_correct, count , color = city
),
size = 0.5) +
ggplot2::scale_y_continuous(
breaks = scales::pretty_breaks(8),
labels = scales::number_format(accuracy = 1)
) +
ggplot2::labs(
title = "Twitter keyword popularity tracking",
caption = paste(
"Note: Hourly data (UTC time) fetched from",
format(as.Date(sinceInput[1]), "%A, %d-%b. %Y") ,
"until",
format(as.Date(Sys.Date()), "%A, %d-%b. %Y"),
".",
sep = " "
),
x = "Days",
y = "Number of tweets posted"
) +
ggplot2::theme_minimal() +
ggplot2::theme(
axis.title = ggplot2::element_text(size = 12, face = "bold"),
plot.caption = ggplot2::element_text(
size = 10,
face = "italic",
hjust = 0
),
plot.title = ggplot2::element_text(size = 14, face = "bold"),
axis.text.x = ggplot2::element_text(
size = 8,
angle = 25,
vjust = 1.0,
hjust = 1.0
),
axis.text.y = ggplot2::element_text(size = 8)
) +
ggplot2::theme(plot.margin = ggplot2::unit(c(2, 10, 2, 2), "mm")) +
viridis::scale_color_viridis(
discrete = TRUE ,
direction = -1,
alpha = 0.6
) +
ggplot2::scale_x_datetime(
labels = scales::date_format("%b %d - %H:%M"),
breaks = scales::date_breaks("5 hour")
) +
ggrepel::geom_text_repel(
data = finaldata %>%
dplyr::group_by(city, created_correct) %>%
dplyr::count() %>%
dplyr::arrange(desc(n)) %>%
dplyr::filter(created_correct == lubridate::today()) %>%
dplyr::rename(count = n) %>%
dplyr::top_n(length(keyword)) ,
ggplot2::aes(x = created_correct,
count , label = factor(city))
,
fontface = 'bold',
color = 'black',
box.padding = 1,
point.padding = 1,
segment.color = 'black' ,
size = 4
) + ggplot2::guides(color = ggplot2::guide_legend(title = "Place"))
)
} else {
finaldata$created <-
lubridate::ymd_hms(finaldata$created,
tz = Sys.timezone(),
quiet = TRUE) %>% lubridate::as_date()
finaldata$created_correct <-
as.POSIXct(format(finaldata$created, "%Y-%m-%d"), format =
"%Y-%m-%d")
print(
finaldata %>%
dplyr::group_by(city, created_correct) %>% dplyr::count() %>%
dplyr::arrange() %>%
dplyr::rename(count = n) %>%
ggplot2::ggplot() +
ggplot2::geom_point(
mapping = ggplot2::aes(x = created_correct, count , color = city),
size = 3,
alpha = 0.8
) +
ggplot2::geom_line(ggplot2::aes(
x = created_correct, count , color = city
),
size = 0.5) +
ggplot2::scale_y_continuous(
breaks = scales::pretty_breaks(8),
labels = scales::number_format(accuracy = 1)
) +
ggplot2::labs(
title = "Twitter keyword popularity tracking",
caption = paste(
"Note: Daily data (UTC time) fetched from",
format(as.Date(sinceInput[1]), "%A, %d-%b. %Y") ,
"until",
format(as.Date(Sys.Date()), "%A, %d-%b. %Y"),
".",
sep = " "
),
x = "Days",
y = "Number of tweets posted"
) +
ggplot2::theme_minimal() +
ggplot2::theme(
axis.title = ggplot2::element_text(size = 12, face = "bold"),
plot.caption = ggplot2::element_text(
size = 10,
face = "italic",
hjust = 0
),
plot.title = ggplot2::element_text(size = 14, face = "bold"),
axis.text.x = ggplot2::element_text(
size = 8,
angle = 25,
vjust = 1.0,
hjust = 1.0
),
axis.text.y = ggplot2::element_text(size = 8)
) +
ggplot2::theme(plot.margin = ggplot2::unit(c(2, 10, 2, 2), "mm")) +
viridis::scale_color_viridis(
discrete = TRUE ,
direction = -1,
alpha = 0.6
) +
ggplot2::scale_x_datetime(
labels = scales::date_format("%b %d"),
breaks = scales::date_breaks("5 hour")
) +
ggrepel::geom_text_repel(
data = finaldata %>%
dplyr::group_by(city, created_correct) %>%
dplyr::count() %>%
dplyr::arrange(desc(n)) %>%
dplyr::filter(created_correct == lubridate::today()) %>%
dplyr::rename(count = n) %>%
dplyr::top_n(length(keyword)) ,
ggplot2::aes(x = created_correct,
count , label = factor(city))
,
fontface = 'bold',
color = 'black',
box.padding = 1,
point.padding = 1,
segment.color = 'black' ,
size = 4
) + ggplot2::guides(color = ggplot2::guide_legend(title = "Place"))
)
#for longer timeframes we aggregate data daily to smooth the
#variations of the dataset
}
#returning the dataframe of the results as well (obtained from Twitter)
return(finaldata)
}
AyrtonRua/group1_project documentation built on May 14, 2019, 5:14 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
#' @title track_twitter_hashtag
#'
#' @description Provide a scatterplot displaying hashtag counts overtime
#' based on data obtained from Twitter.
#'
#' @param keyword A character \code{vector} of maximum length 4
#' (keywords requested)
#' @param type A character \code{vector} of length 1 (type of keyword, can
#' be either a city/place/monument name in English, to be referred to as
#' "place", or if another type of keyword, please specify by "other").
#' \cr
#' Note: all keywords should be of the same types (e.g. all places' names).
#' @param number A positive \code{integer} number (timeframe parameter)
#' @param sincetype A character \code{vector} of length 1 (timeframe parameter),
#' which can be (uniquely) on of the following:
#' either "days", "weeks", "months" or "years"
#' @param provideN A positive \code{integer} number
#' (Number ot tweets to fetch for each keyword)
#'
#' @return A timeseries scatterplot with the hashtag count on the y axis,
#' and on the x axis the timeframe.
#' @return A dataframe containing all the results fetched from Twitter
#' (e.g. tweets) based on package \code{\link{twitteR}}
#'
#' @author Ayrton Rua: \email{ayrton.gomesmartinsrua@unil.ch}
#' @author Maurizio Griffo: \email{maurizio.griffo@unil.ch}
#' @author Ali Karray: \email{mohamedali.karray@unil.ch}
#' @author Mohit Mehrotra: \email{mohit.mehrotra@unil.ch}
#' @author Youness Zarhloul: \email{youness.zarhloul@unil.ch}
#'
#' @references Orso, S., Molinari, R., Lee,
#' J., Guerrier, S., & Beckman, M. (2018).
#' An Introduction to Statistical Programming Methods with R.
#' Retrieved from \url{https://smac-group.github.io/ds/}
#'
#' @seealso \code{\link{touristR}} & \code{\link{run_shiny}}
#'
#' @examples \dontrun{
#' track_keyword(
#' keyword = c("eiffel tower", "san francisco", "london"),
#' type="place",
#' number = 2,
#' sincetype = "weeks",
#' provideN = 100
#' )
#' }
#'
#' @details The function automatically obtains the latitude and longitude of
#' each requested place, then queries the Twitter API to obtain the number of
#' tweets with the requested # keywords for the desired timeframe, and returns
#' at the end a scatterplot based on the
#' aesthetics of \code{\link{ggplot2}},
#' allowing thus to analyze which places are the most popular
#' over a given time period. \cr
#' It should be noted in that sense that if the time frame selected by the user
#' is less than or equal to 2 weeks, the plot returned groups data by hour,
#' whereas if the user requests longer timeframes the result will be grouped by
#' days for better representation of the changes over time
#' (short/long term). \cr \cr
#' Lastly, we advise the user to keep the parameter \code{provideN} low
#' to avoid long delay responses from Twitter API (e.g. setting provideN=100
#' would be preferable than 1000).
#'
#' @importFrom magrittr %>%
#'
#' @export
track_twitter_hashtag <-
function(keyword, type, number, sincetype, provideN) {
#making sure all inputs are correct
stopifnot({
is.vector(keyword, mode = "character")
length(keyword) <= 4
is.vector(type, mode = "character")
length(type) == 1
is.integer(number)
number > 0
is.character(sincetype)
length(sincetype) == 1
sincetype == "days" |
sincetype == "weeks" | sincetype == "months" |
sincetype == "years"
is.integer(provideN)
provideN > 0
})
#accessing Twitter via API
consumer_key <- 'ugjqg8RGNvuTAL1fEiNtw'
consumer_secret <- '4WjuEbP6QLUN2DwDzyTwmdMES6fgnOsS65fWxpT8I'
access_token <-
'76887198-JA3xCVO1vvQMqMDiIobWKKGQxYKSB0CV2lI2PZ7GL'
access_secret <-
'IvzlVOC8KkIaMR5s5K4u2IXbxKQv7EcUSvy2bnaru8gKz'
#
twitteR::setup_twitter_oauth(consumer_key,
consumer_secret,
access_token,
access_secret)
#obtain the latitude and longitude of the keyword (city) input
#and save the result as a dataframe
#obtaining latitude and longitude for each place
df <- data.frame(city = NA,
long = NA,
lat = NA)
if (type == "place") {
for (i in 1:length(keyword)) {
df[i, 1] <- keyword[i]
#obtain latitude and longitude
df[i, 2:3] <- ggmap::geocode(location =
keyword[i] ,
source = "dsk",
messaging = FALSE)
}
} else if (type == "other") {
for (i in 1:length(keyword)) {
df[i, 1] <- keyword[i]
}
}
#vector to save the result (since when date)
sinceInput <- vector(mode = "character", length = 1L)
#adapting the parametersinceInput based on the requested timeframe
if (sincetype == "days") {
sinceInput[1] <-
(lubridate::today(tzone = Sys.timezone()) -
lubridate::days(number)) %>%
lubridate::ymd() %>% as.character()
} else if (sincetype == "weeks") {
sinceInput[1] <-
(lubridate::today(tzone = Sys.timezone()) -
lubridate::weeks(number)) %>% lubridate::ymd() %>% as.character()
} else if (sincetype == "months") {
#time difference in months (respecting e.g. months with 30 and 31 days)
sinceInput[1] <-
(lubridate::today(tzone = Sys.timezone()) -
lubridate::period(1, units = "month")) %>%
lubridate::ymd() %>% as.character()
} else if (sincetype == "years") {
sinceInput[1] <-
(lubridate::today(tzone = Sys.timezone()) -
lubridate::years(number)) %>%
lubridate::ymd() %>% as.character()
}
#saving results of for loop bellow
keyword_input <-
vector(mode = "character", length = length(keyword))
input <-
vector(mode = "character", length = length(keyword))
results_twitter = list() # saving result in a list
#querying Twitter including the geocode
#in order to make sure that the tweets are
#indeed related to the requested place
if (type == "place") {
for (i in 1:length(keyword)) {
#run query for that keyword (format is #keyword without space)
#first we remove any white space to obtain a one word input
keyword_input[i] <-
gsub(keyword[i], pattern = " ", replacement = "")
#then we obtain the #oneword format to be used as input for
#the Twitter query
input[i] <- paste("#", keyword_input[i], sep = "")
#query Twitter using the keyword input
results_twitter[[i]] <- twitteR::searchTwitter(
input[i] ,
n = provideN,
resultType = "mixed",
since = sinceInput[1] ,
until = lubridate::today(tzone = Sys.timezone()) %>%
lubridate::ymd() %>% as.character() ,
#specifying the geocode to be sure we only obtain e.g. indeed
#the results published from Paris for search query Paris (e.g.
#tweet should be made within a radius of 80 miles from Paris maximum)
geocode = paste(df[i, ]$lat, df[i, ]$long, "80mi", sep = ",")
)
} #querying twitter without geocode for type =="other"
} else if (type == "other") {
#query Twitter API
for (i in 1:length(keyword)) {
#run query for that keyword (format is #... without space)
#first we remove any white space to obtain a one word input
keyword_input[i] <-
gsub(keyword[i], pattern = " ", replacement = "")
#then we obtain the #oneword format to be used as input for
#the Twitter query
input[i] <- paste("#", keyword_input[i], sep = "")
#querying Twitter using the keyword input
results_twitter[[i]] <- twitteR::searchTwitter(
input[i] ,
n = provideN,
resultType = "mixed",
since = sinceInput[1] ,
until = lubridate::today(tzone = Sys.timezone()) %>%
lubridate::ymd() %>% as.character()
)
}
}
#obtaining the dataframe with the results fetched from Twitter
#obtaining the results based on the number of keywords requested
#between 1 and 4
if (length(keyword) == 4) {
results_twitter1 <-
results_twitter[[1]] %>% unlist() %>% twitteR::twListToDF() %>%
tibble::as.tibble() %>%
dplyr::mutate(city = keyword[1])
results_twitter2 <-
results_twitter[[2]] %>% unlist() %>% twitteR::twListToDF() %>%
tibble::as.tibble() %>%
dplyr::mutate(city = keyword[2])
results_twitter3 <-
results_twitter[[3]] %>% unlist() %>% twitteR::twListToDF() %>%
tibble::as.tibble() %>%
dplyr::mutate(city = keyword[3])
results_twitter4 <-
results_twitter[[4]] %>% unlist() %>% twitteR::twListToDF() %>%
tibble::as.tibble() %>%
dplyr::mutate(city = keyword[4])
finaldata <-
rbind(results_twitter1,
results_twitter2,
results_twitter3,
results_twitter4)
} else if (length(keyword) == 3) {
results_twitter1 <-
results_twitter[[1]] %>% unlist() %>% twitteR::twListToDF() %>%
tibble::as.tibble() %>%
dplyr::mutate(city = keyword[1])
results_twitter2 <-
results_twitter[[2]] %>% unlist() %>% twitteR::twListToDF() %>%
tibble::as.tibble() %>%
dplyr::mutate(city = keyword[2])
results_twitter3 <-
results_twitter[[3]] %>% unlist() %>% twitteR::twListToDF() %>%
tibble::as.tibble() %>%
dplyr::mutate(city = keyword[3])
finaldata <-
rbind(results_twitter1, results_twitter2, results_twitter3)
} else if (length(keyword) == 2) {
results_twitter1 <-
results_twitter[[1]] %>% unlist() %>%
twitteR::twListToDF() %>% tibble::as.tibble() %>%
dplyr::mutate(city = keyword[1])
results_twitter2 <-
results_twitter[[2]] %>% unlist() %>%
twitteR::twListToDF() %>% tibble::as.tibble() %>%
dplyr::mutate(city = keyword[2])
finaldata <- rbind(results_twitter1, results_twitter2)
} else if (length(keyword) == 1) {
results_twitter1 <-
results_twitter[[1]] %>% unlist() %>%
twitteR::twListToDF() %>% tibble::as.tibble() %>%
dplyr::mutate(city = keyword[1])
finaldata <- results_twitter1
}
#creating the plot of keyword count per day over time
#(measure of popularity)
if (sincetype == "days" && number <= 14) {
#correct date format and aggregate the results by hour which
#can be useful e.g. for marketing campaigns to study engagement
#levels for a certain keyword (e.g. a brand name) thus identifying
#the best time slots to advertise (i.e. when users are interacting
#the most with the brand).
finaldata$created <-
lubridate::ymd_hms(finaldata$created,
tz = Sys.timezone(),
quiet = TRUE) %>%
lubridate::round_date(unit = "hour")
finaldata$created_correct <-
as.POSIXct(format(finaldata$created, "%Y-%m-%d %H:%M:%S"),
format = "%Y-%m-%d %H")
#generating the plot
print(
finaldata %>%
dplyr::group_by(city, created_correct) %>% dplyr::count() %>%
dplyr::arrange() %>%
dplyr::rename(count = n) %>%
ggplot2::ggplot() +
ggplot2::geom_point(
mapping = ggplot2::aes(x = created_correct, count , color = city),
size = 3,
alpha = 0.8
) +
ggplot2::geom_line(ggplot2::aes(
x = created_correct, count , color = city
),
size = 0.5) +
ggplot2::scale_y_continuous(
breaks = scales::pretty_breaks(8),
labels = scales::number_format(accuracy = 1)
) +
ggplot2::labs(
title = "Twitter keyword popularity tracking",
caption = paste(
"Note: Hourly data (UTC time) fetched from",
format(as.Date(sinceInput[1]), "%A, %d-%b. %Y") ,
"until",
format(as.Date(Sys.Date()), "%A, %d-%b. %Y"),
".",
sep = " "
),
x = "Days",
y = "Number of tweets posted"
) +
ggplot2::theme_minimal() +
ggplot2::theme(
axis.title = ggplot2::element_text(size = 12, face = "bold"),
plot.caption = ggplot2::element_text(
size = 10,
face = "italic",
hjust = 0
),
plot.title = ggplot2::element_text(size = 14, face = "bold"),
axis.text.x = ggplot2::element_text(
size = 8,
angle = 25,
vjust = 1.0,
hjust = 1.0
),
axis.text.y = ggplot2::element_text(size = 8)
) +
ggplot2::theme(plot.margin = ggplot2::unit(c(2, 10, 2, 2), "mm")) +
viridis::scale_color_viridis(
discrete = TRUE ,
direction = -1,
alpha = 0.6
) +
ggplot2::scale_x_datetime(
labels = scales::date_format("%b %d - %H:%M"),
breaks = scales::date_breaks("5 hour")
) +
ggrepel::geom_text_repel(
data = finaldata %>%
dplyr::group_by(city, created_correct) %>%
dplyr::count() %>% dplyr::arrange(desc(n)) %>%
dplyr::filter(created_correct == lubridate::today()) %>%
dplyr::rename(count = n) %>%
dplyr::top_n(length(keyword)) ,
ggplot2::aes(x = created_correct,
count , label = factor(city))
,
fontface = 'bold',
color = 'black',
box.padding = 1,
point.padding = 1,
segment.color = 'black' ,
size = 4
) + ggplot2::guides(color = ggplot2::guide_legend(title = "Place"))
)
} else if (sincetype == "weeks" && number < 3) {
#correct date format and aggregate by hour
finaldata$created <-
lubridate::ymd_hms(finaldata$created,
tz = Sys.timezone(),
quiet = TRUE) %>%
lubridate::round_date(unit = "hour")
finaldata$created_correct <-
as.POSIXct(format(finaldata$created, "%Y-%m-%d %H:%M:%S"),
format = "%Y-%m-%d %H")
#plot
print(
finaldata %>%
dplyr::group_by(city, created_correct) %>%
dplyr::count() %>% dplyr::arrange() %>%
dplyr::rename(count = n) %>%
ggplot2::ggplot() +
ggplot2::geom_point(
mapping = ggplot2::aes(x = created_correct, count , color = city),
size = 3,
alpha = 0.8
) +
ggplot2::geom_line(ggplot2::aes(
x = created_correct, count , color = city
),
size = 0.5) +
ggplot2::scale_y_continuous(
breaks = scales::pretty_breaks(8),
labels = scales::number_format(accuracy = 1)
) +
ggplot2::labs(
title = "Twitter keyword popularity tracking",
caption = paste(
"Note: Hourly data (UTC time) fetched from",
format(as.Date(sinceInput[1]), "%A, %d-%b. %Y") ,
"until",
format(as.Date(Sys.Date()), "%A, %d-%b. %Y"),
".",
sep = " "
),
x = "Days",
y = "Number of tweets posted"
) +
ggplot2::theme_minimal() +
ggplot2::theme(
axis.title = ggplot2::element_text(size = 12, face = "bold"),
plot.caption = ggplot2::element_text(
size = 10,
face = "italic",
hjust = 0
),
plot.title = ggplot2::element_text(size = 14, face = "bold"),
axis.text.x = ggplot2::element_text(
size = 8,
angle = 25,
vjust = 1.0,
hjust = 1.0
),
axis.text.y = ggplot2::element_text(size = 8)
) +
ggplot2::theme(plot.margin = ggplot2::unit(c(2, 10, 2, 2), "mm")) +
viridis::scale_color_viridis(
discrete = TRUE ,
direction = -1,
alpha = 0.6
) +
ggplot2::scale_x_datetime(
labels = scales::date_format("%b %d - %H:%M"),
breaks = scales::date_breaks("5 hour")
) +
ggrepel::geom_text_repel(
data = finaldata %>%
dplyr::group_by(city, created_correct) %>%
dplyr::count() %>%
dplyr::arrange(desc(n)) %>%
dplyr::filter(created_correct == lubridate::today()) %>%
dplyr::rename(count = n) %>%
dplyr::top_n(length(keyword)) ,
ggplot2::aes(x = created_correct,
count , label = factor(city))
,
fontface = 'bold',
color = 'black',
box.padding = 1,
point.padding = 1,
segment.color = 'black' ,
size = 4
) + ggplot2::guides(color = ggplot2::guide_legend(title = "Place"))
)
} else {
finaldata$created <-
lubridate::ymd_hms(finaldata$created,
tz = Sys.timezone(),
quiet = TRUE) %>% lubridate::as_date()
finaldata$created_correct <-
as.POSIXct(format(finaldata$created, "%Y-%m-%d"), format =
"%Y-%m-%d")
print(
finaldata %>%
dplyr::group_by(city, created_correct) %>% dplyr::count() %>%
dplyr::arrange() %>%
dplyr::rename(count = n) %>%
ggplot2::ggplot() +
ggplot2::geom_point(
mapping = ggplot2::aes(x = created_correct, count , color = city),
size = 3,
alpha = 0.8
) +
ggplot2::geom_line(ggplot2::aes(
x = created_correct, count , color = city
),
size = 0.5) +
ggplot2::scale_y_continuous(
breaks = scales::pretty_breaks(8),
labels = scales::number_format(accuracy = 1)
) +
ggplot2::labs(
title = "Twitter keyword popularity tracking",
caption = paste(
"Note: Daily data (UTC time) fetched from",
format(as.Date(sinceInput[1]), "%A, %d-%b. %Y") ,
"until",
format(as.Date(Sys.Date()), "%A, %d-%b. %Y"),
".",
sep = " "
),
x = "Days",
y = "Number of tweets posted"
) +
ggplot2::theme_minimal() +
ggplot2::theme(
axis.title = ggplot2::element_text(size = 12, face = "bold"),
plot.caption = ggplot2::element_text(
size = 10,
face = "italic",
hjust = 0
),
plot.title = ggplot2::element_text(size = 14, face = "bold"),
axis.text.x = ggplot2::element_text(
size = 8,
angle = 25,
vjust = 1.0,
hjust = 1.0
),
axis.text.y = ggplot2::element_text(size = 8)
) +
ggplot2::theme(plot.margin = ggplot2::unit(c(2, 10, 2, 2), "mm")) +
viridis::scale_color_viridis(
discrete = TRUE ,
direction = -1,
alpha = 0.6
) +
ggplot2::scale_x_datetime(
labels = scales::date_format("%b %d"),
breaks = scales::date_breaks("5 hour")
) +
ggrepel::geom_text_repel(
data = finaldata %>%
dplyr::group_by(city, created_correct) %>%
dplyr::count() %>%
dplyr::arrange(desc(n)) %>%
dplyr::filter(created_correct == lubridate::today()) %>%
dplyr::rename(count = n) %>%
dplyr::top_n(length(keyword)) ,
ggplot2::aes(x = created_correct,
count , label = factor(city))
,
fontface = 'bold',
color = 'black',
box.padding = 1,
point.padding = 1,
segment.color = 'black' ,
size = 4
) + ggplot2::guides(color = ggplot2::guide_legend(title = "Place"))
)
#for longer timeframes we aggregate data daily to smooth the
#variations of the dataset
}
#returning the dataframe of the results as well (obtained from Twitter)
return(finaldata)
}
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.