Analysis of Social Media: Cluster-Based Product Recommendations

This report explores the use of cluster analysis on social media comments to derive product and service recommendations. The study analyzes 309 random Facebook comments, categorizing them into clusters based on shared traits and sentiments. The methodology involves content analysis, assigning scores to comments based on various parameters such as clarity of ideas, emotional level, and perspective (past, present, future). These scores are then used to form clusters, which are further analyzed using Erikson's stages of Psychosocial Development and Ajzen's theory of planned behavior to understand consumer behavior. The report identifies distinct segments with varying preferences and recommends tailored products and services for each, considering factors like disposable income, social status, and value for money. Limitations of the study include the inability to account for demographic data and the exclusion of non-textual content like images and videos. The overall aim is to provide marketers with insights into leveraging social media data for targeted product placement and promotional strategies, with additional resources available on Desklib.

Analysis of Social Media Comments and
Product/ Service Recommendations by
use of Cluster Analysis
Student Name: Student ID:
Subject Name: Subject ID:
Date Due: Professor Name:
Page | 1

Abstract
Web-based social networking has an awesome impact in our day by day lives. Individuals
share their suppositions, stories, news, and communicate occasions utilizing online networking.
This results in incredible measures of data in online networking. Techniques to compose web-
based social networking presents on help more instructive perspectives of information to clients
are required with the goal that clients can without much of a stretch discover gatherings of posts
that they are keen on. For instance, grouping pertinent subjects together enables business clients
to go straightforwardly to the bunch of business related occasions.
Social media comments analysis works as a useful tool in determining products or
services for marketers. They also benefit is selection of marketing and promotional strategies.
The scope of this paper analysis 309 random comments of individuals and then groups them into
clusters to arrive at product recommendation strategies for them.
Page | 2

Table of Contents
Abstract.......................................................................................................................................................2
1.0 Introduction and Objective..............................................................................................................4
2.0 Method............................................................................................................................................5
3.0 Limitation.........................................................................................................................................7
4.0 Segments Developed.......................................................................................................................8
5.0 Products and Services Recommended for Segments.....................................................................11
6.0 Reference Lists.....................................................................................................................................13
7.0 Appendix..............................................................................................................................................15
Page | 3

1.0 Introduction and Objective
Marketers and companies make use of data of various types in order to promote and
ultimately sell their products1. Social media especially Facebook has become most attractive
platform that giant company’s uses for arriving at marketing and promotional strategies related
to their products. Facebook comments are used by multiple companies with data mining
capabilities, who then conduct various types of analysis on them2. Statistical analysis of
comments from Facebook data from various individuals can reflect trend regarding their
personality, traits, likeliness and various other factors. The data generated can be used by
marketers in making product selection relative to demography and then devise marketing
strategies particular to target that segment3.
The scope of this report has undertaken random sample of 300 comments inclusive of
posts. Though these comments were taken from random samples yet it contained various trends
related to individuals that can be used by marketers. These posts include nature, personality traits
and ideas related to commenters that can be used for developing market segments4. Then clusters
were defined for groups of people, who had similar traits from their comment analysis.
Individual’s distinctive attributes were analysed then products or services were recommended
based on their psyche. The following are the objectives of the study;
1L. Kwok et. al. "Spreading social media messages on Facebook: An analysis of restaurant
business-to-consumer communications."
2Wang et. al., "Gender, topic, and audience response: an analysis of user-generated content on
facebook."
3Dekay et. al., "How large companies react to negative Facebook comments."
4Bortree et. al., "Dialogic strategies and outcomes: An analysis of environmental advocacy
groups’ Facebook profiles."
Page | 4

 Study Objective 1:To understand individual’s Facebook comments on random
topics
 Study Objective 2:To develop deeper understand of Facebook comments by
identifying similar traits
 Study Objective 3:To analyse psyche and distinctive features of groups that
have been clustered
 Study Objective 4:To understand products or service likeliness relative to
cluster groups
2.0 Method
Research was led for this investigation utilizing the system of substance examination.
This investigation inspected the substance of more than 300 heterogeneous comments on
Facebook pages and broke down what sorts of data are being presented by people. Baron Babbie
(2009) in The Act of Social Exploration characterizes content examination as, "the investigation
of recorded human interchanges, for example, books, sites, compositions and laws." Content
examination is an approach to efficiently watch the events of words, expressions, thoughts, or
subjects in composed correspondences (Powell, 2004). To begin with, content examination
empowers scientists to filter through huge volumes of information in a precise and orderly design
Steve Stemler (2001).
Test Choice: For this investigation, the populace for inquire about was characterized as
comments on Facebook pages from different sector of users. These comments were picked as the
Page | 5

focal point of this investigation5. Keeping in mind the end goal to distinguish and group
Facebook comments, a few distinctive hunt systems were connected. The main path was to check
if the comments file from Facebook page gives a connection to its relating clusters. Albeit
conceivably the data on Facebook pages is powerfully changing, data from an earlier day and age
scarcely changes and consequently information covering that era can be considered as steady.
Along these lines, the information was gathered which was data of Facebook posts amid the day
and age. Information gathering comprised of filling helpful data into layouts, taking notes, and
recording related screen captures of the Facebook page6. The initial step was to lead an agenda as
appeared in appendix (Table 2). Posts were sorted into classifications as indicated by their
substance. In the wake of checking on the exploration tests, a few normal subjects or properties
of divider posts was extricated and in this manner utilized for post arrangement in nine
categories (table 2 in appendix). The information gathered about 309 posts was ranked in a scale
of one to seven, where one indicated the lowest level and seven was the highest level of marking.
Likewise, if a post incorporates both a remark and a "like" from a similar client, just the remark
is checked, while remarks or "likes" are not tallied by any stretch of the imagination. All the
posts were then rechecked for anomaly in statements and after agreement of the research team
choice of cluster numbers was finalized7. The analysis of variance for clusters revealed that all of
the nine variables were statistically significant (p value of 0.000) in analysis the clustering.
5Alarcón-del-Amo et. al., "Classifying and profiling social networking site users: A latent
segmentation approach."
6GalvisCarreño et. al., "Analysis of user comments: an approach for software requirements
evolution."
7Gerolimos, "Academic libraries on Facebook: An analysis of users' comments."
Page | 6

Figure 1: Bar Diagram for Final Clusters
3.0 Limitation
Despite the fact that this examination adequately developed a technique for estimating the
utilization of Facebook pages by data clustering from various perspectives, constraints still exist8.
Analysing photographs and recordings might be a decent idea for, since they have a tendency to
draw in more consideration and criticism as exhibited by the outcomes appeared in the tables and
figures. For instance, a notice of a coming occasion can be observed with a review photo of this
occasion rather than only content depictions so the post clustering would be clearer and
conceivably pull in more consideration. When posting joins from different destinations, a brief
and charming remark on the connection can be added to trigger the clients' advantage and
contribution. This technique for look into can't record different sorts of record of web activity or
age of the subjects or gender for that matter. The main information that was utilized is known as
8McCandless et. al.,"Method and system for performing trend analysis of themes in social data."
Page | 7

the "divider", which is presumably the most mainstream Facebook highlight. The cluster analysis
would be more appropriate and significant considering these fields of the subjects.
4.0 Segments Developed
Data of random nature was collected from various samples of Facebook comments and
processed to reveal trends related to person’s individual information. Field information was
collected clarity of ideas, level of emotions, and level of objectivity, past perspective, now
perspective, future perspective, level focus on personalities, level of criticism of corporate
business, level of criticism of government or public services9. All these field data was collected
using score method that was assigned for each comment collected. Score assigned to comments
on the above categories ranged from 1 to 7, with 1 being the least score and 7 being highest
possible score. When cluster centers were grouped according to logical assemblies which
revealed various mean. It indicated that from random samples of 300 people cluster various
trends and personality traits were reflected10. Applying Erikson’s stages of Psychosocial
Development various psychological trends can be understood of persons belonging to specific
cluster group. Homogeneity of categories were observed for cluster value k=3. The above given
bar diagram reflects similarity of individuals when clusters were grouped in k=3.
9Ahuja et. al., "Corporate blogs as tools for consumer segmentation-using cluster analysis for
consumer profiling."
10Bonsón et. al.,"Citizens' engagement on local governments' Facebook sites. An empirical
analysis: The impact of different media and content types in Western Europe."
Page | 8

Figure 2: Erikson's Stages of Psychosocial Development
According to Erikson’s stages of Psychosocial Development it needs to be ascertained
age of the person’s whose comments have been analysed11. It can be said when comments are
analysed that age of individuals must be between 21 to 39 years of age or 40 years to 65 years of
age. As per the given model psychosocial crisis experienced at 21 years to 39 years of age is
related to intimacy or isolation, meaning young adults in their 30s. This group is generally
concerned regarding finding the correct partner and is faced with fear of doing so, such that they
do not have to face life alone12. They need someone with whom they can interact and share
phases of life with. At this stage some chooses to live their lives single, they tend to develop
clarity of ideas. This group have high levels of emotions, they are concerned with their past
perspectives as well as future perspectives. They are more bent on personality trends and are less
11McLeod, "Erik Erikson."
12O Connor, "An analysis of the use of Facebook by international hotel chains."
Page | 9

critical regarding corporates or government13. This group has started earning and has high
amounts of available disposable incomes; hence they are more prone to purchase expensive
products as well as luxury items. They generally prefer fancy products; they desire to purchase
products that would enhance their social status. Products that would make them happy or benefit
them are those that would make them less dependent. A detailed analysis of their products and
service preferences will be undertaken in the next section.
In the second group that is analysed and categorised for the study, their characteristics
features can be classified as generativity versus stagnation. At this stage adults are diagnosed,
who feels more meaning with their work. At this stage, they feel they can contribute to the
society or initiate a change process, and in case they fail they feel to be an unproductive member
of the society. These people generally like value for money products and match their needs.
During this stage most individuals have reached peak in their career, hence they are less willing
to spend extra towards any show-off, they just maintain their social status to whichever segment
they belong to14. This segment of people is generally critical regarding government or public
services and corporate who are unable to provide a match for their products or services. They
generally desire products or services that match their needs well and do not want to spend extra
time in making purchase decision. This segment of the population is more difficult to be made
happy as they are highly critical of products or services. They are merely associated with benefits
arising from a particular product.
Another theory that can reflect regarding consumer behaviour is Icek Ajzen and Martin
Fishbein. The theory focuses importance on existing attitudes of consumers towards decision-
13Sweetser et. al.,"Candidates make good friends: An analysis of candidates' uses of Facebook."
14Wu, et. al., "Understanding customers using Facebook Pages: data mining users feedback
using text analysis."
Page | 10

making process. Most important aspect underlying the theory is existing attitudes of consumers
that make them provide or receive a particular type of behaviour. Analysing clusters according to
theory reveals that;
Cluster 1 that has more consumers, who are focused on the future compared and have better
objective with critical approach. Analysing decision making capabilities of this cluster group it
can be said that they are more likely to be critical regarding various products, having choice that
have connectedness to future utility.
Cluster 2 is less critical in nature and has an objective perspective. This segment of cluster has
critical decision making capability with relatively less understanding related to utility of products
or services.
Cluster 3 behaviour reflects less critical approach with perspective for future. They are less
eager to criticize products and services hence retail products with relative less specification can
be forwarded across to them.
Cluster 4 has again critical set of clients. Decision making characteristics for this group reveals
greater clarity of ideas with a critical perspective with high emotions. Marketers can easily make
use of such data analysis to offer their products accordingly in social media platform in form of
advertisements or whatsoever. This model reflects that Big Data can convert into profits for
companies, which can use these analysis tactics for targeting right customers.
Engel, Kollet, Blackwell (EKB) Model is another model that expands scope of Theory of
Reasoned Action. This model provides a five step process of consumer purchase decision, where
consumer collects data and then processes information for comparing it with expectations.
According to this theory the Clusters cannot be analysed in great detail. Motivation-Need theory
Page | 11

is another critical theory in this domain apart from Hawkins Stern Impulse Buying. In Hawkins
theory of impulse buying, Hawkins Stern highlighted the concept pertain to impulse behaviour in
customers. His analysis was focused on rationale purchasing behaviour as against impulse fit. An
impulse buying was he said, dependent upon external stimuli, which can help analyse cluster for
this Facebook comments. But most accurate theory that can be applied to cluster is Theory of
Reasoned Action.
5.0 Products and Services Recommended for Segments
Analysing comments relative to various segments of the clusters, products and services
recommendations can be developed15. The final cluster developed has four groups revealing
various trends that can be analysed as given below;
Cluster 1: The name for this cluster can be “Critical Thinkers with Future
Perspectives”. Analysing psychographic of this cluster segment is highly critical of government
or public and corporate services. They have perspectives related to future and current state. They
experiences or attaches less importance to emotions and are relatively less focused on
personalities or ideas. This cluster is very active set of people, who aims at contributing to
several issues ranging from government services to public companies, new perspectives, and
future and so on. This cluster group has high ideas that they are eager to express in various
forums. This group can belong to young generation who are employed across various service
categories and have opinions regarding every matter. They will generally be selective and choosy
regarding their product, meaning products with varied genuine attributes will attract them.
15Dashtipour et. al., "Multilingual sentiment analysis: state of the art and independent
comparison of techniques."
Page | 12