github export functions.Rmd

---
title: Functions for the Shiny Application Using Twitter Data
author: Nipunjeet Gujral
data: May 27th, 2018
output: 
  html_document:
    theme: spacelab
    highlight: neon
---

##### Setup
```{r libraries, message=FALSE, warning=FALSE}
# markdown related
library(rmarkdown)

# library to attach to twitter API
library(twitteR)

# library to manipulate data
library(dplyr)
library(stringr)
library(lubridate)
library(tidytext)
library(tm)

# library to visualize data
library(wordcloud)
library(RColorBrewer)
library(plotly)
library(ggplot2)

# shiny related
library(shiny)
library(shinydashboard)
```

```{r authentication, echo=FALSE, message=FALSE, warning=FALSE}
authentication <- function(x){
  consumer_key = "______"
  consumer_secret = "___"
  access_token = "______"
  access_secret = "_____"
  setup_twitter_oauth(consumer_key, consumer_secret, access_token, access_secret)
}
```

```{r froming tweet df, message=FALSE, warning=FALSE, echo=FALSE}
pull_tweets <- function(twitter_handle, n){
      tweets <- userTimeline(user = twitter_handle,                                            # pull tweets
                             n = n, 
                             excludeReplies = TRUE) %>%                                        # source tweets
              twListToDF() %>%                                                                 # reformat into dataframe
              dplyr::select(text, created, retweetCount, favoriteCount, statusSource, id) %>%  # select useful rows
              dplyr::mutate(ratio = retweetCount/favoriteCount,                                # ratio: retweet/favorite
                            new_tz = format(created, tz = "America/Los_Angeles"),              # new time zone
                            day_of_week = wday(new_tz, label = TRUE),                          # day of the week 
                            hour = hour(new_tz),                                               # hour 
                            month = month(new_tz))                                             # month
      return(tweets)
}
```


##### Analysing Tweeting Habits
```{r tweets per day, message=FALSE, warning=FALSE}
# create a bar chart depicting the days tweets are created

tweets_per_day <- function(temp_tweets){
  freq_table <- data.frame(table(temp_tweets$day_of_week)) # forming a dataframe of frequncy per day
  plot <- plot_ly(x = freq_table$Var1,                     # plotting 
                  y = freq_table$Freq,
                  type = "bar") %>% 
    layout(title = "Tweets per Day of the Week")
  
  return(plot)
}
```

```{r tweets per hour, message=FALSE, warning=FALSE}
tweets_per_hour <- function(temp_tweets){
# create a line plot depicting the time at which a tweet was created
  
  freq_table <- data.frame(table(temp_tweets$hour))               # forming data frame of frequency per hour
  plot <- plot_ly(x = freq_table$Var1,                            # plotting 
                  y = freq_table$Freq,
                  type = "scatter",
                  mode = "line+markers") %>% 
      layout(title = "Distribution of Tweets in a Day(PDT Time)",
           xaxis = list(title = "Hour (PDT)"),
           yaxis = list(title = "Frequency"))
  
  return(plot)
}
```

```{r soure piechaart, message=FALSE, warning=FALSE}
tweet_source <- function(temp_tweets){
# create a pie graphs depicting the source from which the tweet was sent 
  
    temp <- str_extract(temp_tweets$statusSource, ">(.*)<")   # selecting all characters between "><"
    temp <- gsub('<', "", temp)                               # removing the "<"
    temp <- gsub('>', "", temp)                               # removing the ">"
    
    temp_df <- data.frame(table(temp))                        # tabling the frequency of sources
    colnames(temp_df) <- c("source", "freq")                  # changing column names accorindingly
     
    pie <- plot_ly(data = temp_df,                            # plotting
                   labels = ~source,
                   values = ~freq,
                   type = "pie",
                   textinfo = 'label+percent',
                   textposition = 'inside')
    return(pie)
}
```


##### Analysing User Engagment
```{r retweet vs favorited, message=FALSE, warning=FALSE}
fav_vs_retweet <- function(temp_tweets) {
# illustrate the point the fact that user engagment mearured via Retweet Count and Favorited Count have a strong, posative correlation
  
  fit <- lm(temp_tweets$retweetCount ~ temp_tweets$favoriteCount)  # linear regression betwenn RT and Fav
  plot <- plot_ly(data = temp_tweets,                              # plotting
                  x = ~favoriteCount,                                               
                  y = ~retweetCount, 
                  type = "scatter") %>% 
    add_lines(x = ~favoriteCount,
              y = fitted(fit)) %>% 
    layout(title = "Retweet vs. Favorited Count",
           xaxis = list(title = "Favorited Count"),
           yaxis = list(title = "Retweet Count"))
  
  return(plot)
}
```

```{r ratio vs created, message=FALSE, warning=FALSE}
ratio_vs_time <- function(temp_tweets){
# illustrate how the twitter API samples tweets according to time
  
  plot <- plot_ly(data = temp_tweets,      # plotting ratio vs. time
               x = ~created, y = ~ratio,
               type = "scatter")
  return(plot)
}
```

```{r retweet and favorited histogramn, message=FALSE, warning=FALSE}
distributions <- function(temp_tweets){
# analyse the distributions of the Retweet Count and Favorited Count metrics
  
  plot <- plot_ly(data = temp_tweets,         # plotting both RT and Fav distribution
             alpha = 0.6) %>%
    add_histogram(x = ~retweetCount, name = "Retweet") %>%
    add_histogram(x = ~favoriteCount, name = "Favorited") %>%
    layout(barmode = "overlay",
           xaxis = list(title = " "),
           title = "Distributions of Retweet and Favorited Counts")
  
  return(plot)
}
```

##### Outputs
```{r Function Calls, message=FALSE, warning=FALSE}
# setup
authentication()
temp_tweets <- pull_tweets("elonmusk", 1000)
View(temp_tweets)
# time analysis
tweets_per_day(temp_tweets)
tweets_per_hour(temp_tweets)
ratio_vs_time(temp_tweets)
# variable distribution
fav_vs_retweet(temp_tweets)
distributions(temp_tweets)
# source
tweet_source(temp_tweets)
```