Learning Social Media Analytics with R

Social media has gained immense popularity and importance so that today almost everyone can't stay away from it. Not only is social media a medium for people to express their views, but also a very powerful tool which can be used by businesses to target new and existing customers and increase revenue. We will discuss some of the main advantages of social media as follows:

The significance and importance of social media is quite evident from the preceding points. In today's interconnected world, social media has almost become indispensable and although it might have a lot of disadvantages, including distractions, if we use it for the right reasons, it can indeed be a very important tool or medium to help us achieve great things.

Even though we have been blowing the trumpet about social media and its significance, I'm sure you are already thinking about pitfalls and disadvantages, which are directly or indirectly caused from social media. We want to cover all aspects of social media including the good and the bad, so let's look at some negative aspects:

We have discussed several pitfalls attached to using social media and some of them are very serious concerns. Proper social media usage guidelines and policies should be borne in mind by everyone because social media is like a magnifying glass: anything you post can be used against you or can potentially prove harmful later. Be it extremely sensitive personal information, or confidential information, like design plans for your next product launch, always think carefully before sharing anything with the rest of the world.

However, if you know what you are doing, social media can definitely be used as a proper tool for your personal as well as professional gain.

We will be analyzing data from diverse social media applications and platforms throughout the course of this book. However, it is essential to have a good grasp of the essential concepts behind any typical analytics process or workflow. While we will be expanding more on data analytics and mining processes later, let us look at a typical social media analytics workflow in the following figure:

From the preceding diagram, we can broadly classify the main steps involved in the analytics workflow as follows:

We will now briefly expand upon each of these four processes since we will be using them extensively in future chapters.

{
"user": {
                    "profile_sidebar_fill_color": "DDFFCC",
                    "profile_sidebar_border_color": "BDDCAD",
                    "profile_background_tile": true,
                    "name": "J'onn J'onzz",
                    "profile_image_url": "http://test.com/img/prof.jpg",
                    "created_at": "Tue Apr 07 19:05:07 +0000 2009",
                    "location": "Ox City, UK",
                    "follow_request_sent": null,
                    "profile_link_color": "0084B4",
                    "is_translator": false,
                    "id_str": "2921138"
},
"followers_count": 2452,
"statuses_count": 7311,
"friends_count": 427
}

Based on the advantages of social media, we can derive plentiful opportunities which lie within the scope of social media analytics. You can save a lot of cost involved in targeted advertising and promotions by analyzing your social media traffic patterns. You can see how users engage with your brand or business using social media, for instance, when it is the perfect time to share something interesting, such as a new service, product, or even an interesting anecdote about your company. Based on traffic from different geographies, you can analyze and understand the preferences of users from different parts of the world. Users love it if you publish promotions in their local language, and businesses are already leveraging such capabilities from social media platforms such as Facebook to target users in specific countries based on localized content.

The social media analytics landscape is still young and emerging and has a lot of untapped potential.

Let us understand the potential of social media analytics better by taking a real-world example.

Consider you are running a profitable business with active engagement on various social media channels. How can you use the data generated from social media to know how you are doing and how your competitors are doing? Live data streams from Twitter could be continuously analyzed to get real-time mood, sentiment, emotion, and reactions of people to your products and services. You could even analyze the same for your rival competitors to see when they are launching their commodities and how users are reacting to them. With Facebook, you can do the same and even push localized promotions and advertisements to see if they help in generating better revenue. News portals would give you live feeds of trending news articles and insights into the current state of the economy and current events and help you decide if these are favorable times for a thriving business or should you be preparing for some hard times. Sentiment analysis, concept mining, topic models, clustering, and inference are just a few examples of using analytics on social media. The opportunities are huge—you just need to have a clear objective in mind so that you can use analytics effectively to solve that objective.

Before we delve into the challenges associated with social media analytics let us look at the following interesting facts:

These statistics give you a rough idea about the massive scale of data being generated and consumed in these social media platforms. This leads to some challenges:

These are perhaps the most prevalent challenges when analyzing social media data, amongst many other challenges, that you might face in your social media analytics journey. Let's now get acquainted with the R programming language, which will be useful to us when we are performing our analyzes.

We will be discussing the necessary steps for setting up a proper analysis environment by installing the necessary dependencies around the R ecosystem and also the necessary code snippets, functions, and modules which we will be using across all the chapters. You can refer to any code snippet being used across any chapter from the code files which will be provided for each chapter along with this book. Besides that, you can also access our GitHub repository https://github.com/dipanjanS/learning-social-media-analytics-with-r for necessary code modules, snippets and functions which will be used in the book and adopt them for your own analyzes!

The R language is free and open-source as we mentioned earlier, and is available for all major operating systems. At the time of writing this book, the latest version of R is 3.3.1 (code named Bug in Your Hair) and is available for downloading at https://www.r-project.org/. This link includes detailed steps, but the direct download page can be accessed at https://cloud.r-project.org/ if you are interested. Download the necessary binary distribution based on your operating system of choice and run the executable setup following the necessary instructions for the Windows platform. If you are using Unix or any *nix like environment, you can install it directly from the terminal too if needed.

Once R is installed, you can fire up the R interpreter directly. This has a graphical user interface (GUI) containing an editor where you can write your code and then execute it. We recommend using an Integrated Development Environment (IDE) instead which eases development and helps maintain code in a more structured way. Besides this you can also use it for other capabilities like generating R markdown documents, R notebooks and Shiny Web Applications. We recommend using RStudio which provides a user-friendly interface for working with R. You can download and install it from https://www.rstudio.com/products/rstudio/download3/ which contains installers for various operating systems.

Once installed, you can start RStudio and use R directly from the IDE itself. It usually contains a code editor window at the top and the R interactive interpreter in the bottom. The interactive interpreter is often called a Read-Evaluate-Print-Loop (REPL). The interpreter asks for input, evaluates and instantly returns the output if any in the interpreter window itself. The interface usually shows the > symbol when waiting for any input and often shows the + symbol in the prompt when you enter code which spans multiple lines. Anything in R is usually a vector and outputs are usually returned with square brackets preceding it, like [1] indicating the output is a vector of size one. Comments are used to describe functions or sections of code. We can specify comments by using the # symbol followed by text. A sample execution in the R interpreter is shown in the following code for convenience:

> 10 + 5
[1] 15
> c(1,2,3,4)
[1] 1 2 3 4
> 5 == 5
[1] TRUE
> fruit = 'apple'
> if (fruit == 'apple'){
+     print('Apple')
+ }else{
+     print('Orange')
+ }
[1] "Apple"

You can see various operations being performed in the R interpreter in the preceding code snippet, including some conditional evaluations and basic arithmetic. We will now delve deeper into the various constructs of R.

The base R system has several important core data structures which are extensively used in handling, processing, manipulating, and analyzing data. We will be talking about five important data structures, which can be classified according to the type of data which can be stored and its dimensionality. The classification is depicted in the following table:

The content type in the table depicts whether the data stored in the structure belongs to the same data type (homogeneous) or can contain data of different data types, (heterogeneous). The dimensionality of the data structure is pretty straightforward and is self-explanatory. We will now examine each data structure in further detail.

> 1:5
[1] 1 2 3 4 5
> c(1,2,3,4,5)
[1] 1 2 3 4 5
> seq(1,5)
[1] 1 2 3 4 5
> seq_len(5)
[1] 1 2 3 4 5

# assigning two vectors to variables
> x <- 1:5
> y <- c(6,7,8,9,10)
> x
[1] 1 2 3 4 5
> y
[1]  6  7  8  9 10

# operating on vectors
> x + y
[1]  7  9 11 13 15
> sum(x)
[1] 15
> mean(x)
[1] 3
> x * y
[1]  6 14 24 36 50
> sqrt(x)
[1] 1.000000 1.414214 1.732051 2.000000 2.236068

# indexing and slicing
> y[2:4]
[1] 7 8 9
> y[c(2,3,4)]
[1] 7 8 9

# naming vector elements
> names(x) <- c("one", "two", "three", "four", "five")
> x
  one   two three  four  five 
    1     2     3     4     5

# create a three-dimensional array
three.dim.array <- array(
    1:12,    # input data
    dim = c(2, 2, 3),   # dimensions
    dimnames = list(    # names of dimensions
        c("row1", "row2"),
        c("col1", "col2"),
        c("first.set", "second.set", "third.set")
    )
)

# view the array
> three.dim.array
, , first.set

     col1 col2
row1    1    3
row2    2    4

, , second.set

     col1 col2
row1    5    7
row2    6    8

, , third.set

     col1 col2
row1    9   11
row2   10   12

# create a matrix
mat <- matrix(
    1:12,   # data
    nrow = 4,  # num of rows
    ncol = 3,  # num of columns
    byrow = TRUE  # fill the elements row-wise
)

# view the matrix
> mat
     [,1] [,2] [,3]
[1,]    1    2    3
[2,]    4    5    6
[3,]    7    8    9
[4,]   10   11   12

# initialize matrices
m1 <- matrix(
    1:9,   # data
    nrow = 3,  # num of rows
    ncol = 3,  # num of columns
    byrow = TRUE  # fill the elements row-wise
)
m2 <- matrix(
    10:18,   # data
    nrow = 3,  # num of rows
    ncol = 3,  # num of columns
    byrow = TRUE  # fill the elements row-wise
)

# matrix addition
> m1 + m2
     [,1] [,2] [,3]
[1,]   11   13   15
[2,]   17   19   21
[3,]   23   25   27

# matrix transpose
> t(m1)
     [,1] [,2] [,3]
[1,]    1    4    7
[2,]    2    5    8
[3,]    3    6    9

# matrix product
> m1 %*% m2
     [,1] [,2] [,3]
[1,]   84   90   96
[2,]  201  216  231
[3,]  318  342  366

# create sample list
list.sample <- list(
    nums = seq.int(1,5),
    languages = c("R", "Python", "Julia", "Java"),
    sin.func = sin
)

# view the list
> list.sample
$nums
[1] 1 2 3 4 5

$languages
[1] "R"      "Python" "Julia"  "Java"  

$sin.func
function (x)  .Primitive("sin")

# accessing individual list elements
> list.sample$languages
[1] "R"      "Python" "Julia"  "Java"  
> list.sample$sin.func(1.5708)
[1] 1

# initializing two lists
l1 <- list(nums = 1:5)
l2 <- list(
    languages = c("R", "Python", "Julia"),
    months = c("Jan", "Feb", "Mar")
)

# check lists and their type
> l1
$nums
[1] 1 2 3 4 5

> typeof(l1)
[1] "list"
> l2
$languages
[1] "R"      "Python" "Julia" 

$months
[1] "Jan" "Feb" "Mar"

> typeof(l2)
[1] "list"

# concatenating lists
> l3 <- c(l1, l2)
> l3
$nums
[1] 1 2 3 4 5

$languages
[1] "R"      "Python" "Julia" 

$months
[1] "Jan" "Feb" "Mar"

# converting list back to a vector
> v1 <- unlist(l1)
> v1
nums1 nums2 nums3 nums4 nums5 
    1     2     3     4     5 
> typeof(v1)
[1] "integer"

# create data frame
df <- data.frame(
  name =  c("Wade", "Steve", "Slade", "Bruce"),
  age = c(28, 85, 55, 45),
  job = c("IT", "HR", "HR", "CS")
)

# view the data frame
> df
   name age job
1  Wade  28  IT
2 Steve  85  HR
3 Slade  55  HR
4 Bruce  45  CS

# examine data frame properties
> class(df)
[1] "data.frame"
> str(df)
'data.frame':      4 obs. of  3 variables:
 $ name: Factor w/ 4 levels "Bruce","Slade",..: 4 3 2 1
 $ age : num  28 85 55 45
 $ job : Factor w/ 3 levels "CS","HR","IT": 3 2 2 1
> rownames(df)
[1] "1" "2" "3" "4"
> colnames(df)
[1] "name" "age" "job"
> dim(df)
[1] 4 3

# initialize two data frames
emp.details <- data.frame(
    empid = c('e001', 'e002', 'e003', 'e004'),
    name = c("Wade", "Steve", "Slade", "Bruce"),
    age = c(28, 85, 55, 45)
)
job.details <- data.frame(
    empid = c('e001', 'e002', 'e003', 'e004'),
    job = c("IT", "HR", "HR", "CS")
)

# view data frames
> emp.details
  empid  name age
1  e001  Wade  28
2  e002 Steve  85
3  e003 Slade  55
4  e004 Bruce  45
> job.details
  empid job
1  e001  IT
2  e002  HR
3  e003  HR
4  e004  CS

# binding and merging data frames
> cbind(emp.details, job.details)
  empid  name age empid job
1  e001  Wade  28  e001  IT
2  e002 Steve  85  e002  HR
3  e003 Slade  55  e003  HR
4  e004 Bruce  45  e004  CS
> merge(emp.details, job.details, by='empid')
  empid  name age job
1  e001  Wade  28  IT
2  e002 Steve  85  HR
3  e003 Slade  55  HR
4  e004 Bruce  45  CS

# subsetting data frame
> subset(emp.details, age > 50)
  empid  name age
2  e002 Steve  85
3  e003 Slade  55

So far we have dealt with various variables and data types and structures for storing data. Functions are just another data type or object in R, albeit a special one which allows us to operate on data and perform actions on data. Functions are useful for modularizing code and separating concerns where needed by dedicating specific actions and operations to different functions and implementing the logic needed for any action inside the function. We will be talking about two types of functions in this section: the built-in functions and the user-defined functions.

> sqrt(7)
[1] 2.645751
> mean(1:5)
[1] 3
> sum(1:5)
[1] 15
> sqrt(1:5)
[1] 1.000000 1.414214 1.732051 2.000000 2.236068
> runif(5)
[1] 0.8880760 0.2925848 0.9240165 0.6535002 0.1891892
> rnorm(5)
[1]  1.90901035 -1.55611066 -0.40784306 -1.88185230  0.02035915

# define the function
square <- function(data){
  return (data^2)
}

# inspect function components
> environment(square)
<environment: R_GlobalEnv>

> formals(square)
$data


> body(square)
{
    return(data^2)
}

# execute the function on data
> square(1:5)
[1]  1  4  9 16 25
> square(12)
[1] 144

When writing complete applications and scripts using R, the flow and execution of code is very important. The flow of code is based on statements, functions, variables, and data structures used in the code and it is all based on the algorithms, business logic, and other rules for solving the problem at hand. There are several constructs which can be used to control the flow of code and we will be discussing primarily the following two constructs:

We will start with looking at various looping constructs for executing the same sections of code multiple times.

# for loop
> for (i in 1:5) {
+     cat(paste(i," "))
+ }
1  2  3  4  5  

> sum <- 0
> for (i in 1:10){
+     sum <- sum + i
+ }

> sum
[1] 55

# while loop
> n <- 1
> while (n <= 5){
+     cat(paste(n, " "))
+     n <- n + 1
+ }
1  2  3  4  5  

# repeat loop
> i <- 1
> repeat{
+     cat(paste(i, " "))
+     if (i >= 5){
+         break  # break out of the infinite loop
+     }
+     i <- i + 1
+ }
1  2  3  4  5  

# using if
> num = 10
> if (num == 10){
+     cat('The number was 10')
+ }
The number was 10

# using if-else
> num = 5
> if (num == 10){
+     cat('The number was 10')
+ } else{
+     cat('The number was not 10')
+ }
The number was not 10

# using if-else if-else
> if (num == 10){
+     cat('The number was 10')
+ } else if (num == 5){
+     cat('The number was 5')
+ } else{
+     cat('No match found')
+ }
The number was 5

# using ifelse(...) function
> ifelse(num == 10, "Number was 10", "Number was not 10")
[1] "Number was not 10"

# using switch(...) function
> for (num in c("5","10","15")){
+     cat(
+         switch(num,
+             "5" = "five",
+             "7" = "seven",
+             "10" = "ten",
+             "No match found"
+     ), "\n")
+ }
five 
ten 
No match found

We can perform several advanced vectorized operations in R, which is useful when dealing with large amounts of data and improves code performance with regards to time taken in executing code. Some advanced constructs in the apply family of functions will be covered in this section, as follows:

Let's look at how each of these functions work in further detail.

# creating a 4x4 matrix
> mat <- matrix(1:16, nrow=4, ncol=4)

# view the matrix
> mat
     [,1] [,2] [,3] [,4]
[1,]    1    5    9   13
[2,]    2    6   10   14
[3,]    3    7   11   15
[4,]    4    8   12   16

# row sums
> apply(mat, 1, sum)
[1] 28 32 36 40
> rowSums(mat)
[1] 28 32 36 40

# row means
> apply(mat, 1, mean)
[1]  7  8  9 10
> rowMeans(mat)
[1]  7  8  9 10

# col sums
> apply(mat, 2, sum)
[1] 10 26 42 58
> colSums(mat)
[1] 10 26 42 58
 
# col means
> apply(mat, 2, mean)
[1]  2.5  6.5 10.5 14.5
> colMeans(mat)
[1]  2.5  6.5 10.5 14.5
 
# row quantiles
> apply(mat, 1, quantile, probs=c(0.25, 0.5, 0.75))
    [,1] [,2] [,3] [,4]
25%    4    5    6    7
50%    7    8    9   10
75%   10   11   12   13

# create and view a list of elements
> l <- list(nums=1:10, even=seq(2,10,2), odd=seq(1,10,2))
> l
$nums
 [1]  1  2  3  4  5  6  7  8  9 10

$even
[1]  2  4  6  8 10

$odd
[1] 1 3 5 7 9

# use lapply on the list
> lapply(l, sum)
$nums
[1] 55

$even
[1] 30

$odd
[1] 25

# create and view a sample list
> l <- list(nums=1:10, even=seq(2,10,2), odd=seq(1,10,2))
> l
$nums
 [1]  1  2  3  4  5  6  7  8  9 10

$even
[1]  2  4  6  8 10

$odd
[1] 1 3 5 7 9

# observe differences between lapply and sapply
> lapply(l, mean)
$nums
[1] 5.5

$even
[1] 6

$odd
[1] 5
> typeof(lapply(l, mean))
[1] "list"

> sapply(l, mean)
nums even  odd 
 5.5  6.0  5.0
> typeof(sapply(l, mean))
[1] "double"

> data <- 1:30
> data
 [1]  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
> groups <- gl(3, 10)
> groups
 [1] 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3
Levels: 1 2 3
> tapply(data, groups, sum)
  1   2   3 
 55 155 255 
> tapply(data, groups, sum, simplify = FALSE)
$'1'
[1] 55

$'2'
[1] 155

$'3'
[1] 255

> list(rep(1,4), rep(2,3), rep(3,2), rep(4,1))
[[1]]
[1] 1 1 1 1

[[2]]
[1] 2 2 2

[[3]]
[1] 3 3

[[4]]
[1] 4

> mapply(rep, 1:4, 4:1)
[[1]]
[1] 1 1 1 1

[[2]]
[1] 2 2 2

[[3]]
[1] 3 3

[[4]]
[1] 4

One of the most important aspects of data analytics is to depict meaningful insights with crisp and concise visualizations. Data visualization is one of the most important aspects of exploratory data analysis, as well as an important medium to present results from any analyzes. There are three main popular plotting systems in R:

The following snippet depicts visualizations using all three plotting systems on the popular iris dataset:

# load the data
> data(iris)

# base plotting system
> boxplot(Sepal.Length~Species,data=iris, 
+         xlab="Species", ylab="Sepal Length", main="Iris Boxplot")

This gives us a set of boxplots using the base plotting system as depicted in the following plot:

# lattice plotting system
> bwplot(Sepal.Length~Species,data=iris, xlab="Species", ylab="Sepal Length", main="Iris Boxplot")

This snippet helps us in creating a set of boxplots for the various species using the lattice plotting system:

# ggplot2 plotting system
> ggplot(data=iris, aes(x=Species, y=Sepal.Length)) + geom_boxplot(aes(fill=Species)) + 
+     ylab("Sepal Length") + ggtitle("Iris Boxplot") +
+     stat_summary(fun.y=mean, geom="point", shape=5, size=4) + theme_bw()

This code snippet gives us the following boxplots using the ggplot2 plotting system:

You can see from the preceding visualizations how each plotting system works and compare the plots across each system. Feel free to experiment with each system and visualize your own data!

This quick refresher in getting started with R should gear you up for the upcoming chapters and also make you more comfortable with the R eco-system, its syntax and features. It's important to remember that you can always get help with any aspect of R in a variety of ways. Besides that, packages in R are perhaps going to be the most important tool in your arsenal for analyzing data. We will briefly list some important commands for each of these two aspects which may come in handy in the future.

Analyzing data is a science and an art. Any analytics process usually has a defined set of steps, which are generally executed in sequence. More than often, analytics being an iterative process, it leads to several of these steps being repeated many times over if necessary. There is an industry standard that is widely followed for data analysis, known as CRISP-DM, which stands for Cross Industry Standard Process for Data Mining. This is a standard data analysis and mining process workflow that describes how to break up any particular data analysis problem into six major stages.

The main stages in the CRISP-DM model are as follows:

The following figure shows the complete flow with the various stages in the CRISP-DM model:

The CRISP-DM model. Source: www.wikipedia.org

In principle, the CRISP-DM model is very clear and concise with straightforward requirements at each step. As a result, this workflow has been widely adopted across the industry and has become an industry standard.

Machine-learning techniques are basically algorithms which can work on data and extract insights from it, which can include discovering, forecasting, or predicting trends and patterns. The idea is to build a model using a combination of data and algorithms which can then be used to work on new, previously unseen data and derive actionable insights.

Each and every technique depends on what type of data it can work on and the objective of the problem we are trying to solve. People often get tempted to learn a couple of algorithms and then try to apply them to every problem. An important point to remember here is that there is no universal machine-learning algorithm which fits all problems. The main inputs to machine-learning algorithms are features which are extracted from data using a process known as feature extraction, which is often coupled with another process called feature engineering (or building new features from existing features). Each feature can be described as an attribute of the dataset, such as your location, age, number of posts, shares and so on, if we were dealing with data related to social media user profiles. Machine-learning techniques can be classified into two major types namely supervised learning and unsupervised learning.

Supervised learning techniques are a subset of the family of machine -learning algorithms which are mainly used in predictive modeling and forecasting. A predictive model is basically a model which can be constructed using a supervised learning algorithm on features or attributes from training data (available data used to train or build the model) such that we can predict using this model on newer, previously unseen data points. Supervised learning algorithms try to model relationships and dependencies between the target prediction output and the input features such that we can predict the output values for new data based on those relationships which it learned from the dataset used during model training or building.

There are two main types of supervised learning techniques

Unsupervised learning techniques are a subset of the family of machine-learning algorithms which are mainly used in pattern detection, dimension reduction and descriptive modeling. A descriptive model is basically a model constructed from an unsupervised machine learning algorithm and features from input data similar to the supervised learning process. However, the output response variables are not present in this case. These algorithms try to use techniques on the input data to mine for rules, detect patterns, and summarize and group the data points which help in deriving meaningful insights and describe the data better to the users. There is no specific concept of training or testing data here since we do not have any specific relationship mapping between input features and output response variables (which do not exist in this case). There are three main types of unsupervised learning techniques: