Online social network analysis with machine learning techniques

TRIBHUWAN UNIVERSITY
INSTITUTE OF ENGINEERING
PASCHIMANCHAL CAMPUS
“ Online Social Network Analysis
Using Machine Learning Techniques ”
PRESENTED BY:
HARI K.C.
072/ MSCK /R/403
Department Of Electronics and Computer Engineering
A Final Project Presentation
On
1

Introduction
 Social Networking (Twitter)
 Natural Language Processing
 Machine Learning
 Opinion Mining
 Comparison and Popularity Scoring
2

Problem Statement
 Use of Internet ,website and social networking is in rise.
 Social media , a source of large mix type and unstructured
information.
 Difficult for the people to get the efficient , reliable and
information in less time.
 Problems while making the decision .
 This Project compare and predict based on user opinion for
effective buying decision and saves time. 3

Objectives
 To determine the features and opinion sentiments
using Natural Language Processing.
 To determine accuracy of Machine Learning
Classifiers.
 To compare the popularity of two Smartphone.
4

Social Networks
 A network of social actors with interaction.
 People, Organization and Celebrities create their profiles and
share information.
 Different information about products, events, politics, stocks
and so on.
 Twitter is third most popular popular social networking sites
with 310,000,000 visitors.
 Released in July,2006 A.D.
 Real time tweets are posted up to 140 characters.
Source: WWW.ebizma.com (The 15 most popular social networking sites, Nov 26,2016)5

Natural Language Processing
 A field of Artificial Intelligence, focus on developing system to
allow computer- human communication using natural
language.
 It makes the computer to understand and process natural
language .
 There are various aspects of Natural Language Processing
such as  Tokenization
 Part of Speech Tagging
 Term frequency-Independent Document Frequency
 Text Classification
6

Machine Learning
 A construction and study of system that can learn from
data.
 Deals with representation and generalization.
 Different classification and clustering algorithms.
 Supervised Machine learning technique - Classification
 Unsupervised Machine learning technique – Clustering
 Classifiers
 Naive Bayes
 Support Vector
 Logistic Regression
7

Opinion Mining
 A study and analysis of people’s opinions, sentiments, evaluations,
appraisals towards various products, services and organization.
 It focus mainly on Positive and Negative opinions.
 3 levels of Opinion Mining
 Document level
 Sentence Level
 Aspect level
 Opinion Mining is a Natural language Processing problem.
 Opinion is Quintuple. O=(E,A,S,H,T)
 Types of Opinions
 Regular Opinions .Eg: The picture quality of Iphone7 plus
is great.
 Comparative Opinions. Eg: An Android Phone User are
more happier than IPhone User.
8

Popularity Prediction
 An ability of Framework to forecast popularity of some
online content.
Expansion of Web and Social Networking sites allows
fast spreading of information around the globe.
Predicting Popularity is the recent research trend
E.g: Predicting US election win 2016.
Predicting Online News Popularity
Predicting weather
 Popularity Based on Features.
Metrics used for prediction is Message characteristics
such as sentiment index and relative strength.
9

Methodology
Figure1: Block diagram of Project system
10

Methodology
continue…..
 Feature Extraction using Twitter API
 Feature Selection using Natural Language Processing
 Classification includes Machine Learning Technique
 Polarity identification using supervised method
 Popularity score obtained using sentiment index and
Relative strength
11

Methodology
i) Sentiment Differences= (Number of Positive Tweet – Number of Negative Tweet)
, Total number of Tweets
ii) Ratio Comparision, Rp= Number of Positive Tweets
Total Number of Tweets
Rn = Number of Negative Tweets
Total Number of Tweets
iii) Sentiment Index I(sentiment) = [Sentiment Differences + 0.5 ] * 100%
2
iv) Post Rate Rate(post)= Number of Total Tweets
Time of sample window
12

Results
Source: https://twitter.com/@SamsungMobile
Source: https://twitter.com/@Iphone_New
13

Results
1) Tweets as extracted from twitter using Twitter Application Programming Interface.
.
2)Tweets features for analysis
14

Results
3)Training sets and Testing sets sample
4) Output tweets with sentiment value and confidence level
15

Results
5) Analyzed Tweets with required features
16

Results
6) Classifiers accuracy
17

Results
6) Classifiers accuracy
18

Results
8) Most Important Features
20

Results
8) Histogram Plots and Line Graph
a) Android SmartPhone
21

Results
a) Android SmartPhone
22

Results
a) IOS Iphone SmartPhone
23

Results
a) IOS Iphone SmartPhone
24

Results
9) Statistical values:
25

Results
27Android Sentiment index=55.1% IOS Sentiment index=44.9%

Results
28Relative strength of Android is 40.8% which is greater than that of IOS(33.3%)

References
[1] A. Agarwal, B. Xie, L. O. Rumbhu and R. Passonnea, "Sentiment Analysis of Twitter Data," in Workshop on
Languages in Social Media,Computational Linguistic Association, 2011.
[2] B. Liu, Sentiment Analysis and Opinion Mining, Morgan and Claypool Publishers, May ,2012.
[3] C. Trasousas, M. Virrou, K. J. Espinoa and D. Carro, "Sentiment Analysis of Facebook Status using Naive Bayes
Classifier for Language Learning," in DOI:10.11.1109/IISA .6623713, 2013.
[4] L. Bing, K. C. Chan and C. Ou, "Public Sentiment Analysis in Twitter Data for Prediction of Company Stock
Price Movements," in DOI: 10.1109/ICEBE, IEEE, 2014.
[5] F. Gelli, T. Urichio, M. Bertini, A. D. Bimbo and S. Fuchang, "Image Popularity Prediction in Social Media
using Sentiment and Context Features," in ACM, 2015.
[6] A. Ghuibi, S. Mohammed and S. Alshomrani, "A Comprehensive survey on web content Extraction Algorithms
and Techniques," in ICISA, 2013.
[7] N. Kaji and M. Kitsureguewa, "Building Lexicon for sentiment analysis from massive collection of html
documents," in Conference on Empirical Methods in NLP(EMNLP-CONLL), 2007.
[8] K. Lee, A. Agarwal and A. Chaudhary, "Real Time Disease Survillence using Twitter Data," in
DOI:10.1145/2487575, 19th ACM SIGKDD, ICKDDM, 2014.
[9] J. A. P and V. D. Katkar, "Sentiment analysis of twitter data using data mining," in International conference on
Information Processing, 2015.
[10] H. Schoen, D. G. Avello, P. T. Metaxas, E. M. M. Strohmaier and P. Glorr, "The Power of Prediction with Social
Media," in Internet Research,VOl.23 Iss: 5 pp 528-543 ,10.1108/IntR, 2013.
[11] S. Shahhedari, H. and M. R. Bin Daud, "Twitter Sentiment Mining: A multi Domain Analysis," in CSIS, 7th
International Conference,IEEE, 2013.
[12] K. Shimada, S. Insoune, N. Maeda and J. Endo, "Analysing Tourism on Twitter for a Local City.," in DOI:
10.1109/SSNE, 2011. 29

Online social network analysis with machine learning techniques

More Related Content

What's hot (20)

Similar to Online social network analysis with machine learning techniques (20)

Recently uploaded (20)

Online social network analysis with machine learning techniques