Hifza been done through taking different measuresHifza been done through taking different measures


Hifza Afzal

Computer Science,
Balochistan University of information technology, engineering, and management
sciences, Quetta, Pakistan. Quetta, Pakistan

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!

order now

[email protected]


Social media is increasingly broadening up its horizons by making possible
different ways of communication between individual and groups. These different
ways make us possible to understand one’s views, behavior, reactions, and
activities. Communication on Facebook is the common way to exchange the
different views. BUITEMS university students show their reactions on different
public Facebook pages through likes. The first classical signal ‘Like’ is the
common feedback expression on Facebook. The main objective of this paper is to
evaluate the reactions of different type of users on different posts. Secondly,
the impact of likes on a different type of posts. This has been done through
taking different measures on the dataset fetched from Facepager tool. The paper
findings reveal that Facebook ‘like’ reaction can be very helpful for
understanding students and their interests as well, in finding the moderate and
weak correlation and regression measures between a type of posts and likes.

Keywords— Social computing, social media, Post likes analysis, Facebook

I.     Introduction

the past years, social media has played a highly influential role in
influencing public opinion globally. This includes, for example, expressing
their satisfaction, happiness, anger, and disapproval textually or visually
1. Facebook has millions of active user groups sharing different views
globally about different topics 2. A large amount of social feedback
expressed by social signals (e.g. like +1, rating) are assigned to web
resources. ‘Like’ is the most common classical signal used to express reaction
towards posts social signal. Conducting analysis on users’ likes in order to
determine their views through understanding their reactions towards different

II.    Proposed approach

BUITEMS University,
having several pages with majority students following and participating in it.
Understanding their views regarding different issues is of core importance.
Finding meanings from their reactions on public participation will be done and
then classifying them distinctively on bases of their likes on different posts.
This will help to understand students’ different perspectives and activities by
deciphering meanings from their likes. This would be done through following

1.Capturing and storing public data from social
media (Facebook).

2.Classifying individuals based on social
media behavior.

3.Analyzing social media users’ likes, such as
their level of            reactions.

4.Predicting future activities based on social
media data.

5.Understanding public perceptions and views.

applicable sponsor/s here. If no
sponsors, delete this text box (sponsors).


paper has been organized in following order: Section III is based on component
details of our thesis. Section IV provides a complete methodology about the analysis
performed on Facebook likes. Section V shows the concluded results based on our
analysis. In Section VI some predictable future works are estimated. Section
VII concludes this paper based on our analysis.

III.   components

The dataset of 75 posts and its likes are obtained and is
used then in Rstudio for further analysis. Results have been obtained in the
form of charts, graphs, and tables.

A.    Dataset collection tool

Facepager is a dataset obtaining tool for social media public
pages. For analysis of likes Facepager (Version 3.7) has been used, to collect
Facebook posts and there likes.

B.    Dataset

The dataset used for analysis of Facebook post likes are
fetched from BUITEMS University public Facebook page (BUITEMS Social Corner).
The page current posts and its likes are obtained from Facepager tool through
selecting maximum pages option consisting of 75 posts. The dataset is exported in
.csv file, further used in Rstudio.

C.    Rstudio

An open-source free IDE used in this paper for statistical
computing on obtained dataset to perform several analyses resulting in several
answers obtained in the form of charts and graphs.

D.   Post types

For performing analysis, posts have been categorized into two
forms, that are video posts and without video or other posts as this will make
us easy to understand and evaluate the interests of students towards these two
different categories of posts. Through this analysis, we will find the effect
of number of likes on these two different types of posts.


IV.   Methodology

A.    Facebook Reaction Analysis

This section discusses the process of analysis performed on
75 posts obtained from BUITEMS page. As, we have seen that last 75 current
posts of page have been fetched along with its likes, which then separated in
two categories including just video posts, the videos that have been shared on
a page and the other remaining posts or non-video posts, for a better
understanding of obtained likes on these categorized posts. Figure 2 shows the
total likes on posts.


Figure 1


The above figure plotted
through ggplot shows the number of likes against each post, where x-axis shows
the post id’s and the y-axis shows the total likes obtained by each post. By
looking at this plot we found that the maximum posts were liked in the range of
65 to 75 and minimum were around 5 to 16.

Further, we also analyzed
the names of people who liked the posts against each post and categorized
posts. The figure 3 shows collectively the names of people who liked posts
along with the percentage of how many posts have they liked.

Figure 2


Figure 3


This part of analysis
shows the interests of users’ participation and reaction ranging from the most
to least active users. These charts have been fetched in order to gain a proper
and unambiguous information regarding every post and percentage of people likes
on it. For the sake, we next created charts shown in figure 4, of all posts and
percentage of each type of post that is shared for how much time.


Figure 4

Figure 5


Figure 4 shows the
partitions of each type of post along with its percentage for example here
BUITEMS videos that are shared contains 8 percent of posts out of all posts and
cricketer’s post of 2 percent. Now it’s time to finally see and analyze the
percentage of people that have liked each type of post. Figure 5 shows, each
type post liked by percentage of people. If we take the same example we see
that BUITEMS video posts were liked by 4 percent of people and cricketer’s post
by 6 percent of people.

Figure 6


From the figure 5, we can
clearly see and analyze the interests and reactions of people towards a
different type of posts. This understanding will also help us in an improvement
of posts and can also help us in monitoring the activities that students are
more likely to be involved in.

B.    Correlation

Finding correlations
between different elements of a dataset is of prior importance as, it shows us
a rate of values that correlate either resulting in strong, moderate or weak
correlation. We then take in count the strong and moderate correlations more
than weak ones. As, in the political communication of Barak Obama’s page during
his election run, his page videos failed to engage users resulting in negative
correlation . Where on the other side the Facebook pages of different brands
showed the neutral statistic of videos shared . The effects of different
posts types were analyzed resulting in the findings that Status posts caused the
greatest number of comments, while videos caused the most likes, proving the
positive correlation . What seems safest to conclude that sentiment may
differ across sites.

As we discussed the two
categories of posts that were made on bases of there types. Video posts were
separated from other remaining posts to easily monitor the reaction of students
towards a different type of posts. The analysis which we performed on these
types are in between (1) all Post ids and likes that video post obtained, (2) Post
ids and the likes that other posts got and in between both (1) and (2). The
results that we obtained after applying correlation on these three cases are
shown in figure 6.


Correlation Table

Other posts and videos



Likes in video posts



Likes in other posts



Figure 7


Our analysis was
conducted by Pearson’s method which allowed us to show the relationships
between our test cases that we provided. The obtained results are justified on
bases of plots of every case we define. Figure 7 shows the moderate correlation
of (0.3987511) between posts and video likes. By looking to figure 8 proves the
weak correlation of (0.05023071) between all posts and non-video post likes.
The correlation between both video and non-video post likes resulted in
moderate correlation (0.5318082) in figure 9.

Figure 8


Figure 9


Figure 10


We have evaluated our
results through boxplot between video and non-video posts figure 10 clearly
shows the similar results found in correlation. The final results of
correlation measures show neutral correlations, where the moderate results are
more effective in our case that the weak correlation.


Figure 11



C.    Regression

The objective of this evaluation is to describe regression
analysis in the context of estimating the relationships among different
independent and dependent values used for forecasting the future results based
on past and present values. Our objectives are based on likes analysis of
posts, where posts are independent and the likes that video post and other
remaining post gets are dependent and we will be analyzing both of these values
to explain the future change, that the change in posts either bring any
variation or not. Figure 11 shows the regression between posts that lies on the
y-axis and video likes lies on the x-axis with an intercept of (511.84225) and
the standard error of (20.94). Where, in figure 12 the results obtained by the
total likes on the x-axis and non-video posts like for intercept and standard
error are (480.10873) and (1.981), respectively.


Figure 12


Figure 13


The data in figure 11
have a linear component that can be described by a best-fit line having a
non-zero slope, showing that some random components cause them to be scattered
somewhat around that best fit line. Where, in figure 12, we clearly can see
that in the given dataset the X-axis as no significant effect on Y-axis proving
that a very less effect can be seen on post likes during the change in posts.
The results generated are easy to interpret as figure 11 and figure 12 explains
the similar results of that correlation with moderate and weak correlations in
them, respectively.


V.    Results

In this paper we
evaluated the 75 posts of Facebook BUITEMS public page and its likes in
different ways in order to get ideas not only about the impact of posts and
there likes on each other but also on users, which in our case are university

A.    User interest towards posts

In figure 5 we analyzed
the concern of students towards several types of posts with the help of likes
that showed their level of interests. This way we found the most, least and
moderate likes on a different type of posts by students, making us understand their
choices regarding type or category of posts that what sort of activities they
prefer more which further will be helpful to take care of their interests.

B.    Correlation and regression
among posts and likes

We performed both the
correlation and regression tests on some specific columns of our dataset coming
up with very similar results, proved that our findings were taken carefully as
well can nearly be accurate. We performed these analyses on posts and there
likes in both. We found moderate correlation among post and video post likes
and weak correlation among post and the other non-video post likes. Now, the
results of regression justify the results obtained during correlation and
confirm the relationship amongst them. The results of regression have the same
regression lines, post and video post likes have nearly positive regression
line whereas, the regression line between posts and non-video post likes has no
or very weak relationship. Finally, these results helped us to decipher the
meanings of several consequences.

VI.   Future work

In this paper, we present
our results from the evaluation of the effect of the post characteristics:
type, the interaction level in terms of a number of likes. For future work, we
can estimate the impact of these reactions in sentiment detection through
comments as well through different “emojis”. Our findings were based on two
categories of post types, for further analysis, posts can be categorized
further. The results presented in this paper are limited to the dataset
obtained from only one Facebook page. In order to confirm our findings, the
scale of dataset can be expanded to perform analysis to the posts gathered from
other BUITEMS Facebook pages as well.


VII.  Conclusion

In this paper, we present
our results from the evaluation of the effect of the post characteristics:
type, the interaction level in terms of a number of likes. For future work, we
can estimate the impact of these reactions in sentiment detection through
comments as well through different “emojis”. Our findings were based on two
categories of post types, for further analysis, posts can be categorized
further. The results presented in this paper are limited to the dataset
obtained from only one Facebook page. In order to confirm our findings, the
scale of dataset can be expanded to perform analysis to the posts gathered from
other BUITEMS Facebook pages as well.