Social Media Analysis for Customer Preferences

This project focuses on analyzing social media data to understand customer preferences and sentiments. It includes two case studies utilizing text mining and sentiment analysis techniques to derive insights from shared articles across various channels. The findings aim to enhance marketing strategies and brand awareness by leveraging customer sentiment data.

Customer Analytics with Social Media
Assignment 1: Social Media Analysis for Understanding Customer
Preferences and Sentiments
(BUS5CA)
By
<Student Name>
(18752031)
La Trobe Business School
Melbourne, Australia

Table of Contents
1. Introduction 1
2. First Case Study 1
2.1 Data Exploration: Article properties 1
2.1.1 Statistics of Shares 1
2.1.2 Characteristics of high number of shares 7
2.2 Keyword analysis using SAS Text Miner 11
3. Second Case Study 15
References 17
Appendix A

List of Figures
Figure 1 Overall Shares vs Title length...........................................................................................8
Figure 2 Overall shares vs Content length......................................................................................9
Figure 3 Overall shares vs Published in weekends..........................................................................9
Figure 4 Topics of whole data vs Worth.......................................................................................15
Figure 5 qplot of Emotions............................................................................................................16

List of Tables
Table 1 Data channel “Lifestyle”: Top 10 Shares 2
Table 2 “Data channel “Entertainment”: Top 10 Shares 2
Table 3 “Data channel “Business”: Top 10 Shares 3
Table 4 “Data channel “social media”: Top 10 Shares 4
Table 5 “Data channel “technology”: Top 10 Shares 5
Table 6 “Data channel “world”: Top 10 Shares 6
Table 7 Channel “Lifestyle”: Top 10 topics 11
Table 8 Channel “Entertainment” : Top 10 topics 12
Table 9 Channel “Business”: Top 10 topics 12
Table 10 Channel “Social Media”: Top 10 topics 13
Table 11 Channel “Technology: Top 10 topics 13
Table 12 channel “World”: Top 10 topics 14
Table 13 “Complete” data: Top 10 topics 14
Table 14 Eight Emotions Counts 16

1. Introduction
Text mining and sentiment analysis, helps to covert unstructured data to meaningful insights, and can be
explained as computation of polarity of emotions and views, and is most popular method for language
processing (Nikhil et al., 2015). For the case study one, our main aims are the detection of the most
popular articles shared on the social media and their characteristics use text mining method. This analysis
would help to leveragethe power of content and using planned social media marketing, this can assist in
sharp rise in customer base and audience. For case study 2, we will mainly focus to extract
emotions/sentiments from the movie review tweets using dictionary based sentiment analytics method.
For sentiment analysis, the process includes different stages like identification of polarity (i.e. positive,
negative, or neutral) of text and extracting the emotions or sentiments (Ghiassi et al., 2016). This analysis
would provide an insight to build better brand awareness. And marketing strategy by gauging attitudes of
customers.
2. First Case Study
1.1 Data Exploration: Article properties
Under life style channel, number of record is 2100, under entertainment channel, number of record is
7059, under business channel, number of record is 6259, under social media channel, number of record is
2325, under technology channel, number of record is 7345, and under world channel, number of record is
8425. But, no associations with channels were observed for 6147 records.
1.1.1 Statistics of Shares
Highest number of shares i.e. 843300 were observed for the article “Leaked: More Low-Cost iPhone
Photos”, this article not associated with previously mentioned six channels.
1

Lifestyle
Under lifestyle channel, highest number of shares i.e. 208300 were observed for the article ‘Obama to
Discuss NSA Reform with Lawmakers’. Table 1 contains top ten shares under lifestyle channel.
Table 1 Data channel “Lifestyle”: Top 10 Shares
Article Title Number of
shares
‘Obama to Discuss NSA Reform With Lawmakers’ 208300
‘No Movie Trailer Is Complete Without This One Line’ 196700
‘87% of American Teenagers Send Text Messages Each Month’ 139600
‘High-Tech Wristband Monitors Mood’ 81200
‘22 Books for Your Ultimate Summer Reading List’ 73100
‘Finalists Exhibit Tech for $465 Million Virtual Border Fence’ 56000
‘Cybersecurity Experts Will Face Off in Mock NetWars’ 54900
‘84% of Smartphone Owners Use Apps While Getting Ready in the
Morning’
54200
‘It's Still Easy to Get Away With Revenge Porn’ 49700
‘Beats SoloA Headphones Sound Great, But You're Paying for Fashion
[REVIEW]’
45100
Entertainment
Under entertainment channel, highest number of shares i.e. 210300 were observed for the article ‘Sprint's
New Plans Guarantee Unlimited Data for Life’. Table 2 contains top ten shares under entertainment
channel.
2

Table 2 “Data channel “Entertainment”: Top 10 Shares
Article Title Number of
shares
‘Sprint's New Plans Guarantee Unlimited Data for Life’ 210300
‘What to Do With Your New Xbox One’ 197600
‘McDonalds Kills Site That Advised Employees to Eat Healthy Meals’ 193400
‘How a $6,000 Video Got 6 Million Views and Launched a Business’ 138700
'Flappy Bird Typing Tutor' Is Even More Frustrating Than the Original’ 112600
‘Russian Hackers Used Microsoft Bug to Spy on Ukraine and NATO’ 109500
‘Lookout Fights Back Against Smartphone Thieves’ 109100
‘Australian Patient Tests Negative for Ebola’ 98500
‘Facebook Makes Inroads in Russia With Yandex Partnership’ 98000
‘Samsung Trial Jury to Apple: Go After Google’ 96000
Business
Under business channel, highest number of shares i.e. 690400 were observed for the article ‘Dove
Experiment Aims to Change the Way You See Yourself’. Table 3 contains top ten shares under business
channel.
Table 3 “Data channel “Business”: Top 10 Shares
Article Title Number of shares
‘Dove Experiment Aims to Change the Way You See Yourself' 690400
3

‘Kanye West Lectures at Harvard About Creativity' 652900
‘It's Hot as Hell in Australia Right Now' 310800
‘BlackBerry Sold 1 Million BlackBerry 10 Smartphones in Q4' 306100
‘IBM Brings Watson to the Masses and Other News You Need to Know' 298400
‘All The Christmas Movies You Need in One Mashup [VIDEO]' 158900
‘Can Beautiful Design Make Your Resume Stand Out?' 144400
‘Apple to Return $32.5 Million for Accidental App Purchases' 139500
‘How Big Data Is Influencing Hiring' 110200
‘MapBox Enables Amazing Custom Maps for Sites and Apps' 106400
Social media
Under social media channel, highest number of shares i.e. 122800 were observed for the video ‘World's
First Sprout-Powered Battery Just Lit Up a Christmas Tree [VIDEO]”’. Table 4 contains top ten shares
under social media channel.
Table 4 “Data channel “social media”: Top 10 Shares
4

Article Title Number of shares
‘World's First Sprout-Powered Battery Just Lit Up a Christmas Tree [VIDEO]' 122800
‘Facebook Hashtags Not Open to Advertising a Yet' 59000
‘5 Fascinating Facts We Learned From Reddit This Week' 57600
‘The Most Memorable Brand Wins and Fails of 2013' 57000
‘Even Superman Can't Resist Photoshopping His Pics [SUNDAY COMIC]' 54100
‘Watch the YouTube Music Awards Here [VIDEO]' 53100
‘The 100 Most Beautiful Songs in the World, According to Reddit' 51900
‘Rebooted NASA Spacecraft Begins a New Mission 36 Years After Launch' 47700
'Doctor Who' Crash Course: 12 Essential Episodes [VIDEOS]' 47400
‘The Tech Behind the Film 'Gravity'’ 41900
Technology
Under technology channel, highest number of shares i.e. 663600 were observed for the article ‘Startup
stories from early hires”’. Table 5 contains top ten shares under technology channel.
Table 5 “Data channel “technology”: Top 10 Shares
5

Article Title Number of shares
‘Startup stories from early hires' 663600
‘Visit 23 Museums and Zoos Free With Google Field Trip App' 104100
‘Instagram Acquires Video Filter App Luma' 96100
‘Kiev Riots at Fever Pitch: The Fiery Scene in 30 Photos' 88500
‘Amazon Drops Kindle Fire Prices For Cyber Monday' 83300
‘Learn to Code for Free With These 10 Online Resources' 71800
‘Hilarious HelloFlo Ad Is the Only Good Thing About Periods [VIDEO]' 70200
‘16 Crucial Things J.K. Rowling Reveals in New Harry Potter Story' 67800
‘Google Wants You to Live 170 Years' 55200
‘All Eyes on Malala Yousafzai on Second International Day of the Girl' 53200
World
Under world channel, highest number of shares i.e. 284700 were observed for the article ‘U.S. Will Now
Monitor All Travelers From Ebola Zone for 21 Days’. Table 6 contains top ten shares under world
channel.
6

Table 6 “Data channel “world”: Top 10 Shares
Article Title Number of
shares
‘U.S. Will Now Monitor All Travelers From Ebola Zone for 21 Days' 284700
‘Study: 54% of Online Adults Would Spend Tax Refunds on Travel' 141400
‘Apple Fixes FaceTime Bug With iOS 7.0.4' 128500
‘Mystery drones fly over French nuclear sites' 115700
‘Prince Harry Reaches South Pole After 200-Mile Trek for Charity' 111300
‘12 Hours in 'Utopia': On the set of Fox's newest reality show' 108400
‘These Glasses Let You Play in 3D Virtual Worlds [VIDEO]' 96500
‘Thousands of Children Orphaned by West Africaas Ebola Crisis' 84800
‘With Lima climate talks entering critical period, Kerry tries to rally leaders to
act'
75500
‘11 People Who Should Cancel Their Gym Memberships' 69300
1.1.2 Characteristics of high number of shares
7

We have analyzed the top 10 shared articles (among total 843300 articles), 51.1 is the mean title length,
4307.4 is the mean content length, and among top 10, 9 articles were published during weekdays. For the
highest shared article, title length has come out to be 35, content length has come out to be 2073, and this
was also published on a weekday.
Overall number of shares vs title lengths has been shown in Figure 1, overall number of shares vs content
lengths has been shown Figure 2, and comparison of overall number of shares vs published on the
weekend has been shown in Figure 3.
It is observed that title length of highly shared articles are primarily falling within 40 and 70, content
length of highly shared articles are primarily less than 6000, and most of the shares (80%) were observed
for the articles published on weekdays.
Figure 1 Overall Shares vs Title length
8