SlideShare a Scribd company logo

Analyzing-Threat-Levels-of-Extremists-using-Tweets

In the age of social media communication, it is easy to modulate the minds of users and also instigate violent actions being taken by them in some cases. There is a need to have a system that can analyze the threat level of tweets from influential users and rank their Twitter handles so that dangerous tweets can be avoided going public on Twitter before fact-checking which can hurt the sentiments of people and can take the shape of violence. The study aims to analyse and rank twitter users according to their influential power and extremism of their tweets to help prevent major protests and violent events. We scraped top trending topics and fetched tweets using those hashtags. We propose a custom ranking algorithm which considers source based and content based features along with a knowledge graph which generates the score and rank the twitter users according to the scores. Our aim with this study is to identify and rank extremist twitter users with regards to their impact and influence. We use a technique that takes into consideration both source based and content-based features of tweets to generate the ranking of the extremist twitter users having a high impact factor

1 of 6
Download to read offline
Analyzing Threat Levels from Tweets of Extremist
Politicians
Divya Pandey
MT20128
divya20128@iiitd.ac.in
Waquar Shamsi
MT20073
waquar20073@iiitd.ac.in
Shubham Bhansali
MT20105
shubham20105@iiitd.ac.in
Kunal Anand
2018293
kunal18293@iiitd.ac.in
Reshan Faraz
Phd19006
reshanf@iiitd.ac.in
ABSTRACT
In the age of social media communication, it is easy to
modulate the minds of users and also instigate violent
actions being taken by them in some cases. There is a need
to have a system that can analyze the threat level of tweets
from influential users and rank their Twitter handles so
that dangerous tweets can be avoided going public on
Twitter before fact-checking which can hurt the sentiments
of people and can take the shape of violence. The study
aims to analyse and rank twitter users according to their
influential power and extremism of their tweets to help
prevent major protests and violent events. We scraped top
trending topics and fetched tweets using those hashtags.
We propose a custom ranking algorithm which considers
source based and content based features along with a
knowledge graph which generates the score and rank the
twitter users according to the scores. Our aim with this
study is to identify and rank extremist twitter users with
regards to their impact and influence. We use a technique
that takes into consideration both source based and
content-based features of tweets to generate the ranking of
the extremist twitter users having a high impact factor.
1. INTRODUCTION
The rapid growth of the internet has brought many changes
in the sharing and obtaining of information with the in-
crease in social media use by active users, and social media
has become the most famous manifesto in the 21st century.
In today’s world, when there are several options to express
ourselves or know information on a particular topic, which is
an advantage if we see from one side; however from the other
side, there is a disadvantage that people can easily get ma-
nipulated by reading false claims or news, and it has created
opportunities to monitor the users. One of the social media
is Twitter. Twitter is a micro-blogging social networking
service where users can put short messages referred to as
tweets that can be shared and liked. Twitter can barter not
just the tweet messages but also the dynamics of political
misconduct. Twitter plays a vital role in understanding the
political alignment of users. For example, it helps us to know
the right-wing extremist and left-wing extremist. Knowing
that the politicians have progressively used Twitter to show
their trust in an ongoing political affair, Twitter has recently
become political advocacy in spreading political advocacy.
The algorithm design and implementation help identify and
rank the Twitter users who are extremists and whose threat
level can impact heavily to society. For this custom rank-
ing algorithm with the idea of a knowledge graph has been
used. This considers both source-based and content-based
features.
Figure 1: Interaction of users on Twitter
2. LITERATURE REVIEW
In [1], the author took 863 Twitter handles and categories
into eight different types based on their role in society. Now
they try to find the interaction between the accounts with
the help of a directed graph where each node represents an
account, and each edge represents the interaction based on
the retweets, comments, and tweets related to an article
that mentions some account node. The authors concluded
the result by mentioned how different Twitter Handles en-
gaged different communities based on their popularity. The
result shows that tweet of an official news channel is trusted
by people. Results also show that tweets involved in debate
primarily engaged with the same geographical area tweet
handles. However, the accounts of the official news outlets
and the deception of those outlets were ignored which is a
limitation.
In [2], the researchers have provided numerous methods to
understand the right and left-leaning of political alignment
and recognize websites frequently visited and tweeted by
Twitter users. They collected tweets of three months dur-
ing the 2010 U.S. mid-term elections and by making use of
different approaches identification of political hashtags was
done. First, the researchers did two kinds of feature anal-
ysis: content-based feature and network-based feature. For
network analysis, they build two networks based on men-
tion and retweet from Twitter. A force-directed algorithm
was used. They did direct examination by combining infor-
mation which is topological and content data. For content-
based features, they made use of TF-IDF vectors. After
that, they applied linear SVM to restrict the left and right-
leaning users. People who utilized left leaned hashtags ap-
peared with negative, and right leaned appeared with the
positive weight of hashtag coefficient. The researcher culmi-
nated that high comprehension hashtags are highly fertile
in providing political leaning and dictating about the web-
sites which are the most tagged and frequently look in on
by users on Twitter. The ranking list is generated according
to domain popularity tells the websites most frequently vis-
ited by the users of Twitter. The researchers have figured
out that the tweet data generated on Twitter plays a vital
role in shaping political opinions and impacting the user.
However, the proposed idea lacks and limits to some of the
aspects like generalizability of the proposed approaches to
the international level of political consultation and systems
for multiple parties.
In [3], the authors observed that the features based on con-
tent were as important as the features based on the source
on Twitter with respect to credibility; their work remains
limited by human annotation to obtain ground truth.
In [4], authors show that identifying polarizing content on
Twitter based only on the content may generate false posi-
tives; instead, they proposed identifying the radical content
using behavioral and psychological properties as well. The
authors used TF-IDF scores of uni-gram, bi-grams, and tri-
grams and used the word2vec model for word embedding
generation. For extracting the radical language, the authors
calculated TF-IDF scores for every gram and used word em-
bedding for capturing semantic meanings. However, their
work remains limited by different evasion techniques.
In [5], the group of researchers created 2 kinds of fake news
datasets: the first one is Fake-news Dataset which com-
prises six different categories of news and the second one
is Celebrities Dataset. Then the authors developed a classi-
fication model based on several linguistic features like read-
ability, punctuation, complete LIWC, etc., and then trained
the model and analyzed the performance of the model on
different sets of linguistic features. When trained the model
on their respective dataset, the model worked fairly, getting
an accuracy of above 0.5 on most of the features. The best
performing classifier for the FakeNewsAMT dataset was de-
rived from the Readability features and for the Celebrity
dataset, the best performing was derived using the Punc-
tuation features. But, one of the limitation was that when
did the cross-domain analysis on these two datasets, there is
a significant loss in accuracy if compared with the “within-
domain” results.
In [6], the researchers have proposed an approach to seman-
tic categorization of multimedia where they have performed
Entity Linking for text content and extracted Semantic Con-
cepts for visual contents. Eventually, this labeling process
made the analysis easier for them. According to the re-
searchers, the graph-based approaches can find the unseen
relations among the multimedia contents, and therefore they
have investigated the capability of Graph convolutional net-
works (GCNs) for the same contents. They have provided
a graph containing encoded values of blog posts as nodes
and encoded multimodal relations as edges in the form of
an adjacency matrix as the input to the GCNs. After train-
ing, the researchers compared a list of GCNs to Multi-Layer
Perceptron (MLP) with the report containing Precision, Ac-
curacy, Recall, and F1 metrics and in all the metrics, the
proposed approach of GCNs exceeded the baseline by a sig-
nificant margin in the domain of performing the qualitative
analysis of extreme information related to politics. How-
ever, multiple GCNs were compared to MLP, and among
the list of GCNs, GCN-TU outperformed other GCNs as
well as MLP in Accuracy, Recall, and F1 while GCN-TECU
outperformed other GCNs and MLP in Precision.
3. DATASET
Election Data was scrapped from twitter using Tweepy API.
84561 tweets were collected. Firstly, manual annotation of
1068 tweets during phase 1 of project later annotation of
remaining 83493 tweets were done using the classifier model
i.e by making use of Multilevel perceptron during phase 2
and 3 as it performed best.
4. PLAN OF WORK
The project aims to propose a system that can analyze the
impact of extremists on twitter and rank them by making
use of knowledge graph and a custom ranking algorithm
which considers different factors.
• First, to predict the tweets belonging to different cat-
egories, i.e., Neutral, Moderate, and Extreme. The
following steps show the way to generate the classes
for each tweet.
– Downloading of Election Tweets.
– Pre-processing of Tweets to get cleaned Tweets.
– Labelling and Manual annotation of 1068 distinct
Tweets on three categories Neutral, Moderate,
and Extreme.
– Using Word2Vec as feature vector representation
then applying of Baseline Models and training of
model on those Tweets.
– To predict the labels of new Tweets through mod-
els applied and these labels would be further used
for ranking of Twitter handles.
– Obtaining Training and Testing Accuracy.
We use this feature in the Figure 6: Prototype Model
- 2 for ranking the twitter users. To rank the twitter
users following two methods are developed
• Considering the two criteria Content-based and Source-
based, ranking of twitter users is done. The whole plan
of work and system working is shown in figure 2 : Plan
of work: Architecture
Figure 2: Plan of Work : Architecture
5. BASELINE RESULTS
Machine Learning models were applied to predict the labels
after the model being trained on 67% of the data set and
tested on 33% of new tweets. Predicted labels show three
categories of Tweets. The Result Table 1 gives the baseline
results.
5.1 DECISION TREE CLASSIFIER
By making use of the attribute selection measure, it chooses
the best attribute to divide the records. A further attribute
is made as a decision node and breaks the data set into
smaller subsets. This process is followed recursively until
the halting condition comes and a tree is built completely.
The project model is trained on 67% of the data, and for the
rest, 33% testing is being performed. The achieved training
and the testing accuracy is shown in the Result Table 1.
5.2 MULTINOMIAL NAIVE BAYES
It is a probabilistic classifier that assumes that the features
it uses are conditionally independent of each other. Multi-
nomial Naive Bayes makes us understand that each p(fi|c)
is a multinomial distribution. Given some class c to find the
probability of features, let’s say from f1 to fn, then Naive
Bayes holds the following:
The project model is trained on 67% of the data, and for the
rest, 33% testing is being performed. The achieved training
and the testing accuracy are shown in the Result Table 1.
5.3 SUPPORT VECTOR MACHINE
It is one of the most robust supervised learning algorithm
used for regression as well as classification problems both, as
it creates the best decision boundary. It works in two ways,
both linear and non-linear. The two-dimensional linearly
separable data can be separated by a line ax1 − x2 + b = 0.
The project model is trained on 67% of the data, and for the
rest, 33% testing is being performed. The achieved training
and the testing accuracy is shown in the Result Table 1.
5.4 RANDOM FOREST CLASSIFIER
It is a supervised learning algorithm. Random forests gener-
ate decision trees on randomly chosen data samples. Then
it tries to obtain predictions from each tree. After that,
it selects the best possible solution using polling or voting.
The project model is trained on 67% of the data, and for the
rest, 33% testing is being performed. The achieved training,
and the testing accuracy are shown in the Result Table 1.
5.5 MULTI-LEVEL PERCEPTRON
It is a kind of feed forward ANN. It uses multiple level of
perceptron layers. It has activation function which is con-
sidered as threshold. When a particular threshold is reached
the perceptron gets fired. It consists of three layers mainly
input, hidden and output layer. Every node is a perceptron
and gets fired depending on whether it cross the threshold or
not. If yes perceptron it gets fired else doesn’t gets fired.It
makes use of loss function gradient to update parameters as
shown below: This deep learning architecture provided the
highest accuracy after the baseline and thus was used to fur-
ther analysis. The achieved accuracy is shown in the Result
Table 2. Below formula tells to update the parameters by
making use of loss function gradient.
5.6 LOGISTIC REGRESSION
It is a supervised statistical learning technique. It uses the
concept of probability. It uses sigmoid as a cost function
instead of the linear function. This sigmoid function is also
called a logistic function. This logistic function maps the
real value to another value between 0 and 1. The project
model is trained on 67% of the data, and for the rest, 33%
testing is being performed. The achieved training and the
testing accuracy are shown in the Result Table 1.
5.7 CUSTOM RANKING ALGORITHM
The algorithm used takes source based and content based
features as a parameter to generate a score for each user ac-
cording to which they are ranked. Normalization of all fea-
tures is essential to prevent dominance of subset of features
in determining the score. The algorithm gave the following
ranking on the dataset used as depicted in the Figure 6:
Prototype Model-2.
6. PROPOSED METHOD
In this project, we propose to develop a system that could
do ranking of the twitter users based on various algorithms
taking different factor into consideration. Custom ranking
algorithm is applied to generate ranking of most extremist
users by taking the basis and cognizance from knowledge
graph. In this we use some content based and source based
features of tweets to rank the twitter users for their extrem-
ism and impact. The content based features include the
polarity of sentiment and a categorical feature which classi-
fies any tweet as ’Normal’, ’Moderate’ or ’Extremist’. For
this we use supervised algorithms where we trained on the
manually annotated dataset. This helps in understanding
the nature of the tweet. We assign labels as Label 0 for
Neutral, Label 1 for Moderate, and Label 2 for Extreme.
The source based feature include number of followers of the
twitter user and the number of tweets the user made, which
can provide insight on reachability of the user’s content. The
degree of centrality and degree of betweeness are the two ma-
jor parameters taken into consideration from the knowledge
graph.In this project, we have used Twitter as the source of
data collection, and this dataset is prepared by first scrap-
ing the top trends from a website using beautiful soup after
which theses trends are used to fetch tweets using tweepy
library to create the entire dataset. It is very important
to have a good representation and quality of data before
applying the model for analysis. Thus Pre-processing is an
important step that improves the generalizability of a model.
6.1 KNOWLEDGE GRAPH
Knowledge graph contains set of vertices and edges where
data points are interlinked and represents relation and pro-
vides meaningful information. A graph G(V,E) where V rep-
resents vertices i.e Twitter users, E represents the edges i.e
represents the relation based on the mentions of the tweets
by user. The degree of centrality and betweeness central-
ity are the major parameters used. Betweeness centrality
algorithm gauges the shortest path between each and every
pair of nodes. The graph made is directed. This made use
of (Breadth first search) BFS algorithm internally. Score
is calculated for every score.This score is dependent on the
shortest path that goes via the nodes. Higher betweeness
centrality score is obtained when the nodes comes in be-
tween the shortest path. Degree of betweeness is calculated
using:
Figure 3: Betweeness Centrality
Centrality measures highly stressed numbers of paths.Centrality
attempts to discover the most important nodes in a graph.
centrality is calculated using:
Figure 4: Centrality
6.2 PREPROCESSING
• Cleaning of Tweets: Removal of URLs, Special Sym-
bols, Links, Hashtags, Usernames. Twitter original
data contains a lot of HTML entities thus, it is impor-
tant to get rid of them.
• Removal of Punctuation: Removal of unnecessary
commas and symbols is very necessary in order to get
cleaned Tweets. “.”, “,”,”?” are necessary punctuation
and thus required to be retained while others have to
be removed.
• Removal of Duplicates: It is very necessary to re-
move duplicate tweets as the data-set downloaded from
Twitter contains the same retweets multiple times. It
is a very important step to remove duplicate tweets
from the Twitter data-set for further processing.
• Word Tokenization: Before applying any of the mod-
els, it is very important to tokenize the dataset so that
we get the words that make processing easier by deep
learning models.
• Removal of Stop Words: Stop words are commonly
used words such as a, an, in which occurs many times
in a sentence, and such words do not add much mean-
ing to sentences, so it is important to remove such
words before applying deep learning algorithms in fur-
ther stages.
• Lemmatization: Morphological analysis is done by
doing lemmatization by removing inflectional endings.
This is done so that grouping of inflected items can be
analyzed as a single item. For this, Vocabulary is used
in a better way.
Figure 5: Prototype Model - 1
Starting from duplicates 2467 tweets, we were left with 1068
distinct tweets after processing. These tweets were manually
annotated into three categories namely Neutral, Moder-
ate, and Extreme describing the level of extremism of the
tweet.Then Word2Vec feature vector representation is ap-
plied. In this project, we applied numerous Machine Learn-
ing algorithms such as Multinomial Naive Bayes, Logistic
Regression, Support Vector Machine, Decision Tree Clas-
sifier, Random Forest to predict the level of extremism in
any tweet. The training and testing accuracy are depicted
in the below Result Table 1. Label prediction is made by
the models, and later these predicted labels are used in the
ranking algorithm. This feature is then used in the Figure
5: Prototype Model - 3 for generating the ranking of twitter
users.
6.3 RANKING GENERATION
For ranking custom based ranking algorithm is used as shown
in figure 5: Prototype Model-2
It generates scores for each user available in the dataset,
depening upon source-based features i.e., the user and
content-based features i.e., the tweets by the user. The
source based features consist of number of followers The con-
tent based features include polarity of sentiment of tweet and
the level of extremism that the tweet belongs to.
The features are normalized so that one feature does not
dominate over others. Knowledge graph is being constructed
on those tweets dataset. It takes two major parameters
into consideration i.e. degree of centrality and degree of
betweeness. Another major factor taken into consideration
is Tweet Count. Thereafter final score is generated using
all these features and then the ranking is done on the ba-
sis on the decreasing value of score as shown in Figure 5:
Prototype Model - 2. Highest score implies top rank. The
baseline results are shown in Figure 2.
Figure 6: Prototype Model - 2
7. CONCLUSION
The project helped in identifying the most extremist twitter
users with the ranking generated. The ranking got improved
with the cognizance and output from the Knowledge graph
and given to the custom ranking algorithm. The increased
accuracy and precision of ranking was due to the extra num-
ber of parameters included in the algorithm. The idea of
knowledge graph found to be very effective as it helped in
identifying the importance of the node. Between centrality
and degree centrality was major parameter in the analysis of
knowledge graph.The deep learning architecture was proved
to be better in analysis in comparison to the baselines mod-
els. Class prediction using model MLP helped in identify-
ing the categories which provided us the information about
which tweets were moderate, neutral or highly extremist.
The above algorithm is thus effective in doing analysis of
threat level of twitter users during the high impact event.
The UI shows the top ranked extremist twitter users on the
basis of hashtags fetching and its tweets collected, by apply-
ing the custom ranking algorithm. This indicates that these
users are highly responsible for generating the threat level
during elections and have high impact during these times.
8. RESULTS
Figure 7: Ranked Result
9. EVALUATION
The baseline results for five different models are shown below
with their training and testing accuracy.
Baseline Models
Models Training
Accuracy
Testing
Accuracy
Naive Bayes 86.29 73.37
Support Vector Machine 86.29 75.92
Logistic Regression 98.04 77.34
Decision Tree Classifier 98.74 76.49
Random Forest 98.74 76.2
Table 1: Result Table
Application of Deep Learning model Multi-Level perceptron
with improved training and testing accuracy.
Deep Learning Model
Models Training
Accuracy
Testing
Accuracy
Multi-Level Perceptron 99.5 82.4
Table 2: New Result Table
10. FUTURE WORK
• Knowledge graph provides a base which is used to
strengthen its search results. Improve the speed of
generating analysis from Knowledge graph,because de-
creased speed may hamper user experience. As the
knowledge graph is applied on big dataset so it con-
sumes a lot amount of time there is need to improve
and speed up so that it can be prevented from ham-
pering the performance.
• Add more analysis on tweets to produce better results.
More analysis can be done by considering the retweet
just like mentions. Further many different computa-
tions can be done on Knowledge graph like Eigen vec-
tor analysis.
11. REFERENCES
[1] M. Cinelli, S. Cresci, A. Galeazzi, W. Quattrocioc-
chi, and M. Tesconi. The limited reach of fake news
on twitter during 2019 european elections. PloS one,
15(6):e0234689, 2020.
[2] M. D. Conover, B. Goncalves, J. Ratkiewicz, A. Flam-
mini, and F. Menczer. Predicting the Political Alignment
of Twitter Users. In 2011 IEEE Third Int’l Conference
on Privacy, Security, Risk and Trust and 2011 IEEE
Third Int’l Conference on Social Computing, pages 192–
199, Boston, MA, USA, Oct. 2011. IEEE.
[3] A. Gupta and P. Kumaraguru. Credibility ranking of
tweets during high impact events. In Proceedings of the
1st Workshop on Privacy and Security in Online Social
Media - PSOSM ’12, pages 2–8, Lyon, France, 2012.
ACM Press.
[4] M. Nouh, J. R. C. Nurse, and M. Goldsmith. Under-
standing the Radical Mind: Identifying Signals to De-
tect Extremist Content on Twitter. arXiv:1905.08067
[cs, stat], May 2019. arXiv: 1905.08067.
[5] V. Pérez-Rosas, B. Kleinberg, A. Lefevre, and
R. Mihalcea. Automatic Detection of Fake News.
arXiv:1708.07104 [cs], Aug. 2017.
[6] S. Rudinac, I. Gornishka, and M. Worring. Multimodal
Classification of Violent Online Political Extremism
Content with Graph Convolutional Networks. In Pro-
ceedings of the on Thematic Workshops of ACM Multi-
media 2017 - Thematic Workshops ’17, pages 245–252,
Mountain View, California, USA, 2017. ACM Press.

Recommended

FRAMEWORK FOR ANALYZING TWITTER TO DETECT COMMUNITY SUSPICIOUS CRIME ACTIVITY
FRAMEWORK FOR ANALYZING TWITTER TO DETECT COMMUNITY SUSPICIOUS CRIME ACTIVITYFRAMEWORK FOR ANALYZING TWITTER TO DETECT COMMUNITY SUSPICIOUS CRIME ACTIVITY
FRAMEWORK FOR ANALYZING TWITTER TO DETECT COMMUNITY SUSPICIOUS CRIME ACTIVITYcscpconf
 
A Survey Of Collaborative Filtering Techniques
A Survey Of Collaborative Filtering TechniquesA Survey Of Collaborative Filtering Techniques
A Survey Of Collaborative Filtering Techniquestengyue5i5j
 
Text mining on Twitter information based on R platform
Text mining on Twitter information based on R platformText mining on Twitter information based on R platform
Text mining on Twitter information based on R platformFayan TAO
 
Final Poster for Engineering Showcase
Final Poster for Engineering ShowcaseFinal Poster for Engineering Showcase
Final Poster for Engineering ShowcaseTucker Truesdale
 
1 Crore Projects | ieee 2016 Projects | 2016 ieee Projects in chennai
1 Crore Projects | ieee 2016 Projects | 2016 ieee Projects in chennai1 Crore Projects | ieee 2016 Projects | 2016 ieee Projects in chennai
1 Crore Projects | ieee 2016 Projects | 2016 ieee Projects in chennai1crore projects
 
INFORMATION RETRIEVAL TOPICS IN TWITTER USING WEIGHTED PREDICTION NETWORK
INFORMATION RETRIEVAL TOPICS IN TWITTER USING WEIGHTED PREDICTION NETWORKINFORMATION RETRIEVAL TOPICS IN TWITTER USING WEIGHTED PREDICTION NETWORK
INFORMATION RETRIEVAL TOPICS IN TWITTER USING WEIGHTED PREDICTION NETWORKIAEME Publication
 
FAKE NEWS DETECTION WITH SEMANTIC FEATURES AND TEXT MINING
FAKE NEWS DETECTION WITH SEMANTIC FEATURES AND TEXT MININGFAKE NEWS DETECTION WITH SEMANTIC FEATURES AND TEXT MINING
FAKE NEWS DETECTION WITH SEMANTIC FEATURES AND TEXT MININGijnlc
 

More Related Content

What's hot

Odsc 2018 detection_classification_of_fake_news_using_cnn_venkatraman
Odsc 2018 detection_classification_of_fake_news_using_cnn_venkatramanOdsc 2018 detection_classification_of_fake_news_using_cnn_venkatraman
Odsc 2018 detection_classification_of_fake_news_using_cnn_venkatramanvenkatramanJ4
 
Twitter Based Sentiment Analysis of Each Presidential Candidate Using Long Sh...
Twitter Based Sentiment Analysis of Each Presidential Candidate Using Long Sh...Twitter Based Sentiment Analysis of Each Presidential Candidate Using Long Sh...
Twitter Based Sentiment Analysis of Each Presidential Candidate Using Long Sh...CSCJournals
 
IRJET- Tweet Segmentation and its Application to Named Entity Recognition
IRJET- Tweet Segmentation and its Application to Named Entity RecognitionIRJET- Tweet Segmentation and its Application to Named Entity Recognition
IRJET- Tweet Segmentation and its Application to Named Entity RecognitionIRJET Journal
 
IRJET- Fake News Detection and Rumour Source Identification
IRJET- Fake News Detection and Rumour Source IdentificationIRJET- Fake News Detection and Rumour Source Identification
IRJET- Fake News Detection and Rumour Source IdentificationIRJET Journal
 
DETECTION OF FAKE ACCOUNTS IN INSTAGRAM USING MACHINE LEARNING
DETECTION OF FAKE ACCOUNTS IN INSTAGRAM USING MACHINE LEARNINGDETECTION OF FAKE ACCOUNTS IN INSTAGRAM USING MACHINE LEARNING
DETECTION OF FAKE ACCOUNTS IN INSTAGRAM USING MACHINE LEARNINGijcsit
 
SENTIMENT ANALYSIS OF TWITTER DATA
SENTIMENT ANALYSIS OF TWITTER DATASENTIMENT ANALYSIS OF TWITTER DATA
SENTIMENT ANALYSIS OF TWITTER DATAParvathy Devaraj
 
Political prediction analysis using text mining and deep learning
Political prediction analysis using text mining and deep learningPolitical prediction analysis using text mining and deep learning
Political prediction analysis using text mining and deep learningVishwambhar Deshpande
 
News construction from microblogging post using open data
News construction from microblogging post using open dataNews construction from microblogging post using open data
News construction from microblogging post using open dataFrancisco Berrizbeitia
 
IRJET - Political Orientation Prediction using Social Media Activity
IRJET -  	  Political Orientation Prediction using Social Media ActivityIRJET -  	  Political Orientation Prediction using Social Media Activity
IRJET - Political Orientation Prediction using Social Media ActivityIRJET Journal
 
Fake News Detection using Machine Learning
Fake News Detection using Machine LearningFake News Detection using Machine Learning
Fake News Detection using Machine Learningijtsrd
 
IRJET - Socirank Identifying and Ranking Prevalent News Topics using Social M...
IRJET - Socirank Identifying and Ranking Prevalent News Topics using Social M...IRJET - Socirank Identifying and Ranking Prevalent News Topics using Social M...
IRJET - Socirank Identifying and Ranking Prevalent News Topics using Social M...IRJET Journal
 
CATEGORIZING 2019-N-COV TWITTER HASHTAG DATA BY CLUSTERING
CATEGORIZING 2019-N-COV TWITTER HASHTAG DATA BY CLUSTERINGCATEGORIZING 2019-N-COV TWITTER HASHTAG DATA BY CLUSTERING
CATEGORIZING 2019-N-COV TWITTER HASHTAG DATA BY CLUSTERINGijaia
 
Seminar on detecting fake accounts in social media using machine learning
Seminar on detecting fake accounts in social media using machine learningSeminar on detecting fake accounts in social media using machine learning
Seminar on detecting fake accounts in social media using machine learningParvathi Sanil Nair
 
Automatic Hate Speech Detection: A Literature Review
Automatic Hate Speech Detection: A Literature ReviewAutomatic Hate Speech Detection: A Literature Review
Automatic Hate Speech Detection: A Literature ReviewDr. Amarjeet Singh
 
Detection and Analysis of Twitter Trending Topics via Link-Anomaly Detection
Detection and Analysis of Twitter Trending Topics via Link-Anomaly DetectionDetection and Analysis of Twitter Trending Topics via Link-Anomaly Detection
Detection and Analysis of Twitter Trending Topics via Link-Anomaly DetectionIJERA Editor
 
Groundhog day: near duplicate detection on twitter
Groundhog day: near duplicate detection on twitterGroundhog day: near duplicate detection on twitter
Groundhog day: near duplicate detection on twitterDan Nguyen
 
IRJET - Election Result Prediction using Sentiment Analysis
IRJET - Election Result Prediction using Sentiment AnalysisIRJET - Election Result Prediction using Sentiment Analysis
IRJET - Election Result Prediction using Sentiment AnalysisIRJET Journal
 

What's hot (19)

Odsc 2018 detection_classification_of_fake_news_using_cnn_venkatraman
Odsc 2018 detection_classification_of_fake_news_using_cnn_venkatramanOdsc 2018 detection_classification_of_fake_news_using_cnn_venkatraman
Odsc 2018 detection_classification_of_fake_news_using_cnn_venkatraman
 
FAKE NEWS DETECTION PPT
FAKE NEWS DETECTION PPT FAKE NEWS DETECTION PPT
FAKE NEWS DETECTION PPT
 
Twitter Based Sentiment Analysis of Each Presidential Candidate Using Long Sh...
Twitter Based Sentiment Analysis of Each Presidential Candidate Using Long Sh...Twitter Based Sentiment Analysis of Each Presidential Candidate Using Long Sh...
Twitter Based Sentiment Analysis of Each Presidential Candidate Using Long Sh...
 
IRJET- Tweet Segmentation and its Application to Named Entity Recognition
IRJET- Tweet Segmentation and its Application to Named Entity RecognitionIRJET- Tweet Segmentation and its Application to Named Entity Recognition
IRJET- Tweet Segmentation and its Application to Named Entity Recognition
 
IRJET- Fake News Detection and Rumour Source Identification
IRJET- Fake News Detection and Rumour Source IdentificationIRJET- Fake News Detection and Rumour Source Identification
IRJET- Fake News Detection and Rumour Source Identification
 
DETECTION OF FAKE ACCOUNTS IN INSTAGRAM USING MACHINE LEARNING
DETECTION OF FAKE ACCOUNTS IN INSTAGRAM USING MACHINE LEARNINGDETECTION OF FAKE ACCOUNTS IN INSTAGRAM USING MACHINE LEARNING
DETECTION OF FAKE ACCOUNTS IN INSTAGRAM USING MACHINE LEARNING
 
SENTIMENT ANALYSIS OF TWITTER DATA
SENTIMENT ANALYSIS OF TWITTER DATASENTIMENT ANALYSIS OF TWITTER DATA
SENTIMENT ANALYSIS OF TWITTER DATA
 
Political prediction analysis using text mining and deep learning
Political prediction analysis using text mining and deep learningPolitical prediction analysis using text mining and deep learning
Political prediction analysis using text mining and deep learning
 
News construction from microblogging post using open data
News construction from microblogging post using open dataNews construction from microblogging post using open data
News construction from microblogging post using open data
 
IRJET - Political Orientation Prediction using Social Media Activity
IRJET -  	  Political Orientation Prediction using Social Media ActivityIRJET -  	  Political Orientation Prediction using Social Media Activity
IRJET - Political Orientation Prediction using Social Media Activity
 
Fake News Detection using Machine Learning
Fake News Detection using Machine LearningFake News Detection using Machine Learning
Fake News Detection using Machine Learning
 
Seminar Report Mine
Seminar Report MineSeminar Report Mine
Seminar Report Mine
 
IRJET - Socirank Identifying and Ranking Prevalent News Topics using Social M...
IRJET - Socirank Identifying and Ranking Prevalent News Topics using Social M...IRJET - Socirank Identifying and Ranking Prevalent News Topics using Social M...
IRJET - Socirank Identifying and Ranking Prevalent News Topics using Social M...
 
CATEGORIZING 2019-N-COV TWITTER HASHTAG DATA BY CLUSTERING
CATEGORIZING 2019-N-COV TWITTER HASHTAG DATA BY CLUSTERINGCATEGORIZING 2019-N-COV TWITTER HASHTAG DATA BY CLUSTERING
CATEGORIZING 2019-N-COV TWITTER HASHTAG DATA BY CLUSTERING
 
Seminar on detecting fake accounts in social media using machine learning
Seminar on detecting fake accounts in social media using machine learningSeminar on detecting fake accounts in social media using machine learning
Seminar on detecting fake accounts in social media using machine learning
 
Automatic Hate Speech Detection: A Literature Review
Automatic Hate Speech Detection: A Literature ReviewAutomatic Hate Speech Detection: A Literature Review
Automatic Hate Speech Detection: A Literature Review
 
Detection and Analysis of Twitter Trending Topics via Link-Anomaly Detection
Detection and Analysis of Twitter Trending Topics via Link-Anomaly DetectionDetection and Analysis of Twitter Trending Topics via Link-Anomaly Detection
Detection and Analysis of Twitter Trending Topics via Link-Anomaly Detection
 
Groundhog day: near duplicate detection on twitter
Groundhog day: near duplicate detection on twitterGroundhog day: near duplicate detection on twitter
Groundhog day: near duplicate detection on twitter
 
IRJET - Election Result Prediction using Sentiment Analysis
IRJET - Election Result Prediction using Sentiment AnalysisIRJET - Election Result Prediction using Sentiment Analysis
IRJET - Election Result Prediction using Sentiment Analysis
 

Similar to Analyzing-Threat-Levels-of-Extremists-using-Tweets

IRJET- An Experimental Evaluation of Mechanical Properties of Bamboo Fiber Re...
IRJET- An Experimental Evaluation of Mechanical Properties of Bamboo Fiber Re...IRJET- An Experimental Evaluation of Mechanical Properties of Bamboo Fiber Re...
IRJET- An Experimental Evaluation of Mechanical Properties of Bamboo Fiber Re...IRJET Journal
 
Spammer Detection and Fake User Identification on Social Networks
Spammer Detection and Fake User Identification on Social NetworksSpammer Detection and Fake User Identification on Social Networks
Spammer Detection and Fake User Identification on Social NetworksIRJET Journal
 
SENTIMENT ANALYSIS OF TWITTER DATA
SENTIMENT ANALYSIS OF TWITTER DATASENTIMENT ANALYSIS OF TWITTER DATA
SENTIMENT ANALYSIS OF TWITTER DATAanargha gangadharan
 
REAL TIME SENTIMENT ANALYSIS OF TWITTER DATA
REAL TIME SENTIMENT ANALYSIS OF TWITTER DATAREAL TIME SENTIMENT ANALYSIS OF TWITTER DATA
REAL TIME SENTIMENT ANALYSIS OF TWITTER DATAMary Lis Joseph
 
DP1_160430723010_Divya.pptx
DP1_160430723010_Divya.pptxDP1_160430723010_Divya.pptx
DP1_160430723010_Divya.pptxDivyaPatel729457
 
SENTIMENT ANALYSIS OF SOCIAL MEDIA DATA USING DEEP LEARNING
SENTIMENT ANALYSIS OF SOCIAL MEDIA DATA USING DEEP LEARNINGSENTIMENT ANALYSIS OF SOCIAL MEDIA DATA USING DEEP LEARNING
SENTIMENT ANALYSIS OF SOCIAL MEDIA DATA USING DEEP LEARNINGIRJET Journal
 
IRJET- Effective Countering of Communal Hatred During Disaster Events in Soci...
IRJET- Effective Countering of Communal Hatred During Disaster Events in Soci...IRJET- Effective Countering of Communal Hatred During Disaster Events in Soci...
IRJET- Effective Countering of Communal Hatred During Disaster Events in Soci...IRJET Journal
 
IRJET- Sentiment Analysis using Machine Learning
IRJET- Sentiment Analysis using Machine LearningIRJET- Sentiment Analysis using Machine Learning
IRJET- Sentiment Analysis using Machine LearningIRJET Journal
 
IRJET- Improved Real-Time Twitter Sentiment Analysis using ML & Word2Vec
IRJET-  	  Improved Real-Time Twitter Sentiment Analysis using ML & Word2VecIRJET-  	  Improved Real-Time Twitter Sentiment Analysis using ML & Word2Vec
IRJET- Improved Real-Time Twitter Sentiment Analysis using ML & Word2VecIRJET Journal
 
IRJET - Detection of Drug Abuse using Social Media Mining
IRJET - Detection of Drug Abuse using Social Media MiningIRJET - Detection of Drug Abuse using Social Media Mining
IRJET - Detection of Drug Abuse using Social Media MiningIRJET Journal
 
IRJET- Identification of Prevalent News from Twitter and Traditional Media us...
IRJET- Identification of Prevalent News from Twitter and Traditional Media us...IRJET- Identification of Prevalent News from Twitter and Traditional Media us...
IRJET- Identification of Prevalent News from Twitter and Traditional Media us...IRJET Journal
 
Twitter: Social Network Or News Medium?
Twitter: Social Network Or News Medium?Twitter: Social Network Or News Medium?
Twitter: Social Network Or News Medium?Serge Beckers
 
Twitter: Social Network Or News Medium?
Twitter: Social Network Or News Medium?Twitter: Social Network Or News Medium?
Twitter: Social Network Or News Medium?Serge Beckers
 
Twitter Based Election Prediction and Analysis
Twitter Based Election Prediction and AnalysisTwitter Based Election Prediction and Analysis
Twitter Based Election Prediction and AnalysisIRJET Journal
 
Twitter_Hashtag_Prediction.pptx
Twitter_Hashtag_Prediction.pptxTwitter_Hashtag_Prediction.pptx
Twitter_Hashtag_Prediction.pptxSayaliKawale2
 
Measuring information credibility in social media using combination of user p...
Measuring information credibility in social media using combination of user p...Measuring information credibility in social media using combination of user p...
Measuring information credibility in social media using combination of user p...IJECEIAES
 

Similar to Analyzing-Threat-Levels-of-Extremists-using-Tweets (20)

F017433947
F017433947F017433947
F017433947
 
IRJET- An Experimental Evaluation of Mechanical Properties of Bamboo Fiber Re...
IRJET- An Experimental Evaluation of Mechanical Properties of Bamboo Fiber Re...IRJET- An Experimental Evaluation of Mechanical Properties of Bamboo Fiber Re...
IRJET- An Experimental Evaluation of Mechanical Properties of Bamboo Fiber Re...
 
Spammer Detection and Fake User Identification on Social Networks
Spammer Detection and Fake User Identification on Social NetworksSpammer Detection and Fake User Identification on Social Networks
Spammer Detection and Fake User Identification on Social Networks
 
SENTIMENT ANALYSIS OF TWITTER DATA
SENTIMENT ANALYSIS OF TWITTER DATASENTIMENT ANALYSIS OF TWITTER DATA
SENTIMENT ANALYSIS OF TWITTER DATA
 
REAL TIME SENTIMENT ANALYSIS OF TWITTER DATA
REAL TIME SENTIMENT ANALYSIS OF TWITTER DATAREAL TIME SENTIMENT ANALYSIS OF TWITTER DATA
REAL TIME SENTIMENT ANALYSIS OF TWITTER DATA
 
E017433538
E017433538E017433538
E017433538
 
DP1_160430723010_Divya.pptx
DP1_160430723010_Divya.pptxDP1_160430723010_Divya.pptx
DP1_160430723010_Divya.pptx
 
SENTIMENT ANALYSIS OF SOCIAL MEDIA DATA USING DEEP LEARNING
SENTIMENT ANALYSIS OF SOCIAL MEDIA DATA USING DEEP LEARNINGSENTIMENT ANALYSIS OF SOCIAL MEDIA DATA USING DEEP LEARNING
SENTIMENT ANALYSIS OF SOCIAL MEDIA DATA USING DEEP LEARNING
 
IRJET- Effective Countering of Communal Hatred During Disaster Events in Soci...
IRJET- Effective Countering of Communal Hatred During Disaster Events in Soci...IRJET- Effective Countering of Communal Hatred During Disaster Events in Soci...
IRJET- Effective Countering of Communal Hatred During Disaster Events in Soci...
 
IRJET- Sentiment Analysis using Machine Learning
IRJET- Sentiment Analysis using Machine LearningIRJET- Sentiment Analysis using Machine Learning
IRJET- Sentiment Analysis using Machine Learning
 
IRJET- Improved Real-Time Twitter Sentiment Analysis using ML & Word2Vec
IRJET-  	  Improved Real-Time Twitter Sentiment Analysis using ML & Word2VecIRJET-  	  Improved Real-Time Twitter Sentiment Analysis using ML & Word2Vec
IRJET- Improved Real-Time Twitter Sentiment Analysis using ML & Word2Vec
 
IRJET - Detection of Drug Abuse using Social Media Mining
IRJET - Detection of Drug Abuse using Social Media MiningIRJET - Detection of Drug Abuse using Social Media Mining
IRJET - Detection of Drug Abuse using Social Media Mining
 
nm
nmnm
nm
 
IRJET- Identification of Prevalent News from Twitter and Traditional Media us...
IRJET- Identification of Prevalent News from Twitter and Traditional Media us...IRJET- Identification of Prevalent News from Twitter and Traditional Media us...
IRJET- Identification of Prevalent News from Twitter and Traditional Media us...
 
Twitter: Social Network Or News Medium?
Twitter: Social Network Or News Medium?Twitter: Social Network Or News Medium?
Twitter: Social Network Or News Medium?
 
Twitter: Social Network Or News Medium?
Twitter: Social Network Or News Medium?Twitter: Social Network Or News Medium?
Twitter: Social Network Or News Medium?
 
Twitter Based Election Prediction and Analysis
Twitter Based Election Prediction and AnalysisTwitter Based Election Prediction and Analysis
Twitter Based Election Prediction and Analysis
 
Twitter_Hashtag_Prediction.pptx
Twitter_Hashtag_Prediction.pptxTwitter_Hashtag_Prediction.pptx
Twitter_Hashtag_Prediction.pptx
 
Measuring information credibility in social media using combination of user p...
Measuring information credibility in social media using combination of user p...Measuring information credibility in social media using combination of user p...
Measuring information credibility in social media using combination of user p...
 
Kushin (2018) review of Meltwater, Journal of Public Relations Education, Vol...
Kushin (2018) review of Meltwater, Journal of Public Relations Education, Vol...Kushin (2018) review of Meltwater, Journal of Public Relations Education, Vol...
Kushin (2018) review of Meltwater, Journal of Public Relations Education, Vol...
 

More from RESHAN FARAZ

Rm literature survey_on_scheduling_and_resource_allocation_for_ieee_80211ax_p...
Rm literature survey_on_scheduling_and_resource_allocation_for_ieee_80211ax_p...Rm literature survey_on_scheduling_and_resource_allocation_for_ieee_80211ax_p...
Rm literature survey_on_scheduling_and_resource_allocation_for_ieee_80211ax_p...RESHAN FARAZ
 
HR Recruitment System
HR Recruitment SystemHR Recruitment System
HR Recruitment SystemRESHAN FARAZ
 
Stock Price Prediction Using Time Series and Sentiment Analysis
Stock Price Prediction Using Time Series and Sentiment AnalysisStock Price Prediction Using Time Series and Sentiment Analysis
Stock Price Prediction Using Time Series and Sentiment AnalysisRESHAN FARAZ
 
DSA (Data Structure and Algorithm) Questions
DSA (Data Structure and Algorithm) QuestionsDSA (Data Structure and Algorithm) Questions
DSA (Data Structure and Algorithm) QuestionsRESHAN FARAZ
 
Lets crack neet and jee (quick revision)
Lets crack neet and jee (quick revision)Lets crack neet and jee (quick revision)
Lets crack neet and jee (quick revision)RESHAN FARAZ
 
Scheduling and Resource allocation in 802.11ax (WIFI 6)
Scheduling and Resource allocation in 802.11ax (WIFI 6)Scheduling and Resource allocation in 802.11ax (WIFI 6)
Scheduling and Resource allocation in 802.11ax (WIFI 6)RESHAN FARAZ
 
B.tech Final year RKGIT (CSE) Face Detection Attendance System
B.tech Final year RKGIT (CSE) Face Detection Attendance SystemB.tech Final year RKGIT (CSE) Face Detection Attendance System
B.tech Final year RKGIT (CSE) Face Detection Attendance SystemRESHAN FARAZ
 
E-Budgetting : Web Application PPT
E-Budgetting : Web Application PPT E-Budgetting : Web Application PPT
E-Budgetting : Web Application PPT RESHAN FARAZ
 
How to Fight Corruption
How to Fight CorruptionHow to Fight Corruption
How to Fight CorruptionRESHAN FARAZ
 

More from RESHAN FARAZ (12)

Rm literature survey_on_scheduling_and_resource_allocation_for_ieee_80211ax_p...
Rm literature survey_on_scheduling_and_resource_allocation_for_ieee_80211ax_p...Rm literature survey_on_scheduling_and_resource_allocation_for_ieee_80211ax_p...
Rm literature survey_on_scheduling_and_resource_allocation_for_ieee_80211ax_p...
 
HR Recruitment System
HR Recruitment SystemHR Recruitment System
HR Recruitment System
 
Stock Price Prediction Using Time Series and Sentiment Analysis
Stock Price Prediction Using Time Series and Sentiment AnalysisStock Price Prediction Using Time Series and Sentiment Analysis
Stock Price Prediction Using Time Series and Sentiment Analysis
 
DSA (Data Structure and Algorithm) Questions
DSA (Data Structure and Algorithm) QuestionsDSA (Data Structure and Algorithm) Questions
DSA (Data Structure and Algorithm) Questions
 
Lets crack neet and jee (quick revision)
Lets crack neet and jee (quick revision)Lets crack neet and jee (quick revision)
Lets crack neet and jee (quick revision)
 
Scheduling and Resource allocation in 802.11ax (WIFI 6)
Scheduling and Resource allocation in 802.11ax (WIFI 6)Scheduling and Resource allocation in 802.11ax (WIFI 6)
Scheduling and Resource allocation in 802.11ax (WIFI 6)
 
B.tech Final year RKGIT (CSE) Face Detection Attendance System
B.tech Final year RKGIT (CSE) Face Detection Attendance SystemB.tech Final year RKGIT (CSE) Face Detection Attendance System
B.tech Final year RKGIT (CSE) Face Detection Attendance System
 
E-Budgetting : Web Application PPT
E-Budgetting : Web Application PPT E-Budgetting : Web Application PPT
E-Budgetting : Web Application PPT
 
How to Fight Corruption
How to Fight CorruptionHow to Fight Corruption
How to Fight Corruption
 
Plasma physics
Plasma physicsPlasma physics
Plasma physics
 
A.I.PPT
A.I.PPTA.I.PPT
A.I.PPT
 
Tensorflow
TensorflowTensorflow
Tensorflow
 

Recently uploaded

Operations Data On Mobile - inSis Mobile App - Sample Screens
Operations Data On Mobile - inSis Mobile App - Sample ScreensOperations Data On Mobile - inSis Mobile App - Sample Screens
Operations Data On Mobile - inSis Mobile App - Sample ScreensKondapi V Siva Rama Brahmam
 
Unlocking New Insights Into the World of European Soccer Through the European...
Unlocking New Insights Into the World of European Soccer Through the European...Unlocking New Insights Into the World of European Soccer Through the European...
Unlocking New Insights Into the World of European Soccer Through the European...ThinkInnovation
 
ISO 27701:2022 Data Privacy New Version Presentation
ISO 27701:2022 Data Privacy New Version PresentationISO 27701:2022 Data Privacy New Version Presentation
ISO 27701:2022 Data Privacy New Version Presentationyogaallworks
 
A Gentle Introduction to Text Analysis :)
A Gentle Introduction to Text Analysis :)A Gentle Introduction to Text Analysis :)
A Gentle Introduction to Text Analysis :)UNCResearchHub
 
Basics of Creating Graphs / Charts using Microsoft Excel
Basics of Creating Graphs / Charts using Microsoft ExcelBasics of Creating Graphs / Charts using Microsoft Excel
Basics of Creating Graphs / Charts using Microsoft ExcelTope Osanyintuyi
 
Customer Satisfaction Data - Multiple Linear Regression Model.pdf
Customer Satisfaction Data -  Multiple Linear Regression Model.pdfCustomer Satisfaction Data -  Multiple Linear Regression Model.pdf
Customer Satisfaction Data - Multiple Linear Regression Model.pdfruwanp2000
 
fundamentals of digital imaging - POONAM.pptx
fundamentals of digital imaging - POONAM.pptxfundamentals of digital imaging - POONAM.pptx
fundamentals of digital imaging - POONAM.pptxPoonamRijal
 
Choose your perfect jacket.pdf
Choose your perfect jacket.pdfChoose your perfect jacket.pdf
Choose your perfect jacket.pdfAlexia Trejo
 
itc limited word file.pdf...............
itc limited word file.pdf...............itc limited word file.pdf...............
itc limited word file.pdf...............mahetamanav24
 
Introduction to data science.pdf-Definition,types and application of Data Sci...
Introduction to data science.pdf-Definition,types and application of Data Sci...Introduction to data science.pdf-Definition,types and application of Data Sci...
Introduction to data science.pdf-Definition,types and application of Data Sci...DrSumathyV
 
Artificial Intelligence for Vision: A walkthrough of recent breakthroughs
Artificial Intelligence for Vision:  A walkthrough of recent breakthroughsArtificial Intelligence for Vision:  A walkthrough of recent breakthroughs
Artificial Intelligence for Vision: A walkthrough of recent breakthroughsNikolas Markou
 
Tips to Align with Your Salesforce Data Goals
Tips to Align with Your Salesforce Data GoalsTips to Align with Your Salesforce Data Goals
Tips to Align with Your Salesforce Data GoalsDataArchiva
 
Ratio analysis, Formulas, Advantage PPt.pptx
Ratio analysis, Formulas, Advantage PPt.pptxRatio analysis, Formulas, Advantage PPt.pptx
Ratio analysis, Formulas, Advantage PPt.pptxSugumarVenkai
 
Cousera Cap Course Datasets containing datasets from a Fictional Fitness Trac...
Cousera Cap Course Datasets containing datasets from a Fictional Fitness Trac...Cousera Cap Course Datasets containing datasets from a Fictional Fitness Trac...
Cousera Cap Course Datasets containing datasets from a Fictional Fitness Trac...Samuel Chukwuma
 
What you need to know about Generative AI and Data Management?
What you need to know about Generative AI and Data Management?What you need to know about Generative AI and Data Management?
What you need to know about Generative AI and Data Management?Denodo
 
EXCEL-VLOOKUP-AND-HLOOKUP LECTURE NOTES ALL EXCEL VLOOKUP NOTES PDF
EXCEL-VLOOKUP-AND-HLOOKUP LECTURE NOTES ALL EXCEL VLOOKUP NOTES PDFEXCEL-VLOOKUP-AND-HLOOKUP LECTURE NOTES ALL EXCEL VLOOKUP NOTES PDF
EXCEL-VLOOKUP-AND-HLOOKUP LECTURE NOTES ALL EXCEL VLOOKUP NOTES PDFProject Cubicle
 

Recently uploaded (17)

Operations Data On Mobile - inSis Mobile App - Sample Screens
Operations Data On Mobile - inSis Mobile App - Sample ScreensOperations Data On Mobile - inSis Mobile App - Sample Screens
Operations Data On Mobile - inSis Mobile App - Sample Screens
 
Unlocking New Insights Into the World of European Soccer Through the European...
Unlocking New Insights Into the World of European Soccer Through the European...Unlocking New Insights Into the World of European Soccer Through the European...
Unlocking New Insights Into the World of European Soccer Through the European...
 
ISO 27701:2022 Data Privacy New Version Presentation
ISO 27701:2022 Data Privacy New Version PresentationISO 27701:2022 Data Privacy New Version Presentation
ISO 27701:2022 Data Privacy New Version Presentation
 
A Gentle Introduction to Text Analysis :)
A Gentle Introduction to Text Analysis :)A Gentle Introduction to Text Analysis :)
A Gentle Introduction to Text Analysis :)
 
Basics of Creating Graphs / Charts using Microsoft Excel
Basics of Creating Graphs / Charts using Microsoft ExcelBasics of Creating Graphs / Charts using Microsoft Excel
Basics of Creating Graphs / Charts using Microsoft Excel
 
Customer Satisfaction Data - Multiple Linear Regression Model.pdf
Customer Satisfaction Data -  Multiple Linear Regression Model.pdfCustomer Satisfaction Data -  Multiple Linear Regression Model.pdf
Customer Satisfaction Data - Multiple Linear Regression Model.pdf
 
fundamentals of digital imaging - POONAM.pptx
fundamentals of digital imaging - POONAM.pptxfundamentals of digital imaging - POONAM.pptx
fundamentals of digital imaging - POONAM.pptx
 
Choose your perfect jacket.pdf
Choose your perfect jacket.pdfChoose your perfect jacket.pdf
Choose your perfect jacket.pdf
 
itc limited word file.pdf...............
itc limited word file.pdf...............itc limited word file.pdf...............
itc limited word file.pdf...............
 
Introduction to data science.pdf-Definition,types and application of Data Sci...
Introduction to data science.pdf-Definition,types and application of Data Sci...Introduction to data science.pdf-Definition,types and application of Data Sci...
Introduction to data science.pdf-Definition,types and application of Data Sci...
 
Artificial Intelligence for Vision: A walkthrough of recent breakthroughs
Artificial Intelligence for Vision:  A walkthrough of recent breakthroughsArtificial Intelligence for Vision:  A walkthrough of recent breakthroughs
Artificial Intelligence for Vision: A walkthrough of recent breakthroughs
 
Electricity Year 2023_updated_22022024.pptx
Electricity Year 2023_updated_22022024.pptxElectricity Year 2023_updated_22022024.pptx
Electricity Year 2023_updated_22022024.pptx
 
Tips to Align with Your Salesforce Data Goals
Tips to Align with Your Salesforce Data GoalsTips to Align with Your Salesforce Data Goals
Tips to Align with Your Salesforce Data Goals
 
Ratio analysis, Formulas, Advantage PPt.pptx
Ratio analysis, Formulas, Advantage PPt.pptxRatio analysis, Formulas, Advantage PPt.pptx
Ratio analysis, Formulas, Advantage PPt.pptx
 
Cousera Cap Course Datasets containing datasets from a Fictional Fitness Trac...
Cousera Cap Course Datasets containing datasets from a Fictional Fitness Trac...Cousera Cap Course Datasets containing datasets from a Fictional Fitness Trac...
Cousera Cap Course Datasets containing datasets from a Fictional Fitness Trac...
 
What you need to know about Generative AI and Data Management?
What you need to know about Generative AI and Data Management?What you need to know about Generative AI and Data Management?
What you need to know about Generative AI and Data Management?
 
EXCEL-VLOOKUP-AND-HLOOKUP LECTURE NOTES ALL EXCEL VLOOKUP NOTES PDF
EXCEL-VLOOKUP-AND-HLOOKUP LECTURE NOTES ALL EXCEL VLOOKUP NOTES PDFEXCEL-VLOOKUP-AND-HLOOKUP LECTURE NOTES ALL EXCEL VLOOKUP NOTES PDF
EXCEL-VLOOKUP-AND-HLOOKUP LECTURE NOTES ALL EXCEL VLOOKUP NOTES PDF
 

Analyzing-Threat-Levels-of-Extremists-using-Tweets

  • 1. Analyzing Threat Levels from Tweets of Extremist Politicians Divya Pandey MT20128 divya20128@iiitd.ac.in Waquar Shamsi MT20073 waquar20073@iiitd.ac.in Shubham Bhansali MT20105 shubham20105@iiitd.ac.in Kunal Anand 2018293 kunal18293@iiitd.ac.in Reshan Faraz Phd19006 reshanf@iiitd.ac.in ABSTRACT In the age of social media communication, it is easy to modulate the minds of users and also instigate violent actions being taken by them in some cases. There is a need to have a system that can analyze the threat level of tweets from influential users and rank their Twitter handles so that dangerous tweets can be avoided going public on Twitter before fact-checking which can hurt the sentiments of people and can take the shape of violence. The study aims to analyse and rank twitter users according to their influential power and extremism of their tweets to help prevent major protests and violent events. We scraped top trending topics and fetched tweets using those hashtags. We propose a custom ranking algorithm which considers source based and content based features along with a knowledge graph which generates the score and rank the twitter users according to the scores. Our aim with this study is to identify and rank extremist twitter users with regards to their impact and influence. We use a technique that takes into consideration both source based and content-based features of tweets to generate the ranking of the extremist twitter users having a high impact factor. 1. INTRODUCTION The rapid growth of the internet has brought many changes in the sharing and obtaining of information with the in- crease in social media use by active users, and social media has become the most famous manifesto in the 21st century. In today’s world, when there are several options to express ourselves or know information on a particular topic, which is an advantage if we see from one side; however from the other side, there is a disadvantage that people can easily get ma- nipulated by reading false claims or news, and it has created opportunities to monitor the users. One of the social media is Twitter. Twitter is a micro-blogging social networking service where users can put short messages referred to as tweets that can be shared and liked. Twitter can barter not just the tweet messages but also the dynamics of political misconduct. Twitter plays a vital role in understanding the political alignment of users. For example, it helps us to know the right-wing extremist and left-wing extremist. Knowing that the politicians have progressively used Twitter to show their trust in an ongoing political affair, Twitter has recently become political advocacy in spreading political advocacy. The algorithm design and implementation help identify and rank the Twitter users who are extremists and whose threat level can impact heavily to society. For this custom rank- ing algorithm with the idea of a knowledge graph has been used. This considers both source-based and content-based features. Figure 1: Interaction of users on Twitter 2. LITERATURE REVIEW In [1], the author took 863 Twitter handles and categories into eight different types based on their role in society. Now they try to find the interaction between the accounts with the help of a directed graph where each node represents an account, and each edge represents the interaction based on the retweets, comments, and tweets related to an article that mentions some account node. The authors concluded the result by mentioned how different Twitter Handles en- gaged different communities based on their popularity. The result shows that tweet of an official news channel is trusted by people. Results also show that tweets involved in debate primarily engaged with the same geographical area tweet handles. However, the accounts of the official news outlets and the deception of those outlets were ignored which is a limitation. In [2], the researchers have provided numerous methods to
  • 2. understand the right and left-leaning of political alignment and recognize websites frequently visited and tweeted by Twitter users. They collected tweets of three months dur- ing the 2010 U.S. mid-term elections and by making use of different approaches identification of political hashtags was done. First, the researchers did two kinds of feature anal- ysis: content-based feature and network-based feature. For network analysis, they build two networks based on men- tion and retweet from Twitter. A force-directed algorithm was used. They did direct examination by combining infor- mation which is topological and content data. For content- based features, they made use of TF-IDF vectors. After that, they applied linear SVM to restrict the left and right- leaning users. People who utilized left leaned hashtags ap- peared with negative, and right leaned appeared with the positive weight of hashtag coefficient. The researcher culmi- nated that high comprehension hashtags are highly fertile in providing political leaning and dictating about the web- sites which are the most tagged and frequently look in on by users on Twitter. The ranking list is generated according to domain popularity tells the websites most frequently vis- ited by the users of Twitter. The researchers have figured out that the tweet data generated on Twitter plays a vital role in shaping political opinions and impacting the user. However, the proposed idea lacks and limits to some of the aspects like generalizability of the proposed approaches to the international level of political consultation and systems for multiple parties. In [3], the authors observed that the features based on con- tent were as important as the features based on the source on Twitter with respect to credibility; their work remains limited by human annotation to obtain ground truth. In [4], authors show that identifying polarizing content on Twitter based only on the content may generate false posi- tives; instead, they proposed identifying the radical content using behavioral and psychological properties as well. The authors used TF-IDF scores of uni-gram, bi-grams, and tri- grams and used the word2vec model for word embedding generation. For extracting the radical language, the authors calculated TF-IDF scores for every gram and used word em- bedding for capturing semantic meanings. However, their work remains limited by different evasion techniques. In [5], the group of researchers created 2 kinds of fake news datasets: the first one is Fake-news Dataset which com- prises six different categories of news and the second one is Celebrities Dataset. Then the authors developed a classi- fication model based on several linguistic features like read- ability, punctuation, complete LIWC, etc., and then trained the model and analyzed the performance of the model on different sets of linguistic features. When trained the model on their respective dataset, the model worked fairly, getting an accuracy of above 0.5 on most of the features. The best performing classifier for the FakeNewsAMT dataset was de- rived from the Readability features and for the Celebrity dataset, the best performing was derived using the Punc- tuation features. But, one of the limitation was that when did the cross-domain analysis on these two datasets, there is a significant loss in accuracy if compared with the “within- domain” results. In [6], the researchers have proposed an approach to seman- tic categorization of multimedia where they have performed Entity Linking for text content and extracted Semantic Con- cepts for visual contents. Eventually, this labeling process made the analysis easier for them. According to the re- searchers, the graph-based approaches can find the unseen relations among the multimedia contents, and therefore they have investigated the capability of Graph convolutional net- works (GCNs) for the same contents. They have provided a graph containing encoded values of blog posts as nodes and encoded multimodal relations as edges in the form of an adjacency matrix as the input to the GCNs. After train- ing, the researchers compared a list of GCNs to Multi-Layer Perceptron (MLP) with the report containing Precision, Ac- curacy, Recall, and F1 metrics and in all the metrics, the proposed approach of GCNs exceeded the baseline by a sig- nificant margin in the domain of performing the qualitative analysis of extreme information related to politics. How- ever, multiple GCNs were compared to MLP, and among the list of GCNs, GCN-TU outperformed other GCNs as well as MLP in Accuracy, Recall, and F1 while GCN-TECU outperformed other GCNs and MLP in Precision. 3. DATASET Election Data was scrapped from twitter using Tweepy API. 84561 tweets were collected. Firstly, manual annotation of 1068 tweets during phase 1 of project later annotation of remaining 83493 tweets were done using the classifier model i.e by making use of Multilevel perceptron during phase 2 and 3 as it performed best. 4. PLAN OF WORK The project aims to propose a system that can analyze the impact of extremists on twitter and rank them by making use of knowledge graph and a custom ranking algorithm which considers different factors. • First, to predict the tweets belonging to different cat- egories, i.e., Neutral, Moderate, and Extreme. The following steps show the way to generate the classes for each tweet. – Downloading of Election Tweets. – Pre-processing of Tweets to get cleaned Tweets. – Labelling and Manual annotation of 1068 distinct Tweets on three categories Neutral, Moderate, and Extreme. – Using Word2Vec as feature vector representation then applying of Baseline Models and training of model on those Tweets. – To predict the labels of new Tweets through mod- els applied and these labels would be further used for ranking of Twitter handles. – Obtaining Training and Testing Accuracy. We use this feature in the Figure 6: Prototype Model - 2 for ranking the twitter users. To rank the twitter users following two methods are developed
  • 3. • Considering the two criteria Content-based and Source- based, ranking of twitter users is done. The whole plan of work and system working is shown in figure 2 : Plan of work: Architecture Figure 2: Plan of Work : Architecture 5. BASELINE RESULTS Machine Learning models were applied to predict the labels after the model being trained on 67% of the data set and tested on 33% of new tweets. Predicted labels show three categories of Tweets. The Result Table 1 gives the baseline results. 5.1 DECISION TREE CLASSIFIER By making use of the attribute selection measure, it chooses the best attribute to divide the records. A further attribute is made as a decision node and breaks the data set into smaller subsets. This process is followed recursively until the halting condition comes and a tree is built completely. The project model is trained on 67% of the data, and for the rest, 33% testing is being performed. The achieved training and the testing accuracy is shown in the Result Table 1. 5.2 MULTINOMIAL NAIVE BAYES It is a probabilistic classifier that assumes that the features it uses are conditionally independent of each other. Multi- nomial Naive Bayes makes us understand that each p(fi|c) is a multinomial distribution. Given some class c to find the probability of features, let’s say from f1 to fn, then Naive Bayes holds the following: The project model is trained on 67% of the data, and for the rest, 33% testing is being performed. The achieved training and the testing accuracy are shown in the Result Table 1. 5.3 SUPPORT VECTOR MACHINE It is one of the most robust supervised learning algorithm used for regression as well as classification problems both, as it creates the best decision boundary. It works in two ways, both linear and non-linear. The two-dimensional linearly separable data can be separated by a line ax1 − x2 + b = 0. The project model is trained on 67% of the data, and for the rest, 33% testing is being performed. The achieved training and the testing accuracy is shown in the Result Table 1. 5.4 RANDOM FOREST CLASSIFIER It is a supervised learning algorithm. Random forests gener- ate decision trees on randomly chosen data samples. Then it tries to obtain predictions from each tree. After that, it selects the best possible solution using polling or voting. The project model is trained on 67% of the data, and for the rest, 33% testing is being performed. The achieved training, and the testing accuracy are shown in the Result Table 1. 5.5 MULTI-LEVEL PERCEPTRON It is a kind of feed forward ANN. It uses multiple level of perceptron layers. It has activation function which is con- sidered as threshold. When a particular threshold is reached the perceptron gets fired. It consists of three layers mainly input, hidden and output layer. Every node is a perceptron and gets fired depending on whether it cross the threshold or not. If yes perceptron it gets fired else doesn’t gets fired.It makes use of loss function gradient to update parameters as shown below: This deep learning architecture provided the highest accuracy after the baseline and thus was used to fur- ther analysis. The achieved accuracy is shown in the Result Table 2. Below formula tells to update the parameters by making use of loss function gradient. 5.6 LOGISTIC REGRESSION It is a supervised statistical learning technique. It uses the concept of probability. It uses sigmoid as a cost function instead of the linear function. This sigmoid function is also called a logistic function. This logistic function maps the real value to another value between 0 and 1. The project model is trained on 67% of the data, and for the rest, 33% testing is being performed. The achieved training and the testing accuracy are shown in the Result Table 1. 5.7 CUSTOM RANKING ALGORITHM The algorithm used takes source based and content based features as a parameter to generate a score for each user ac- cording to which they are ranked. Normalization of all fea- tures is essential to prevent dominance of subset of features in determining the score. The algorithm gave the following ranking on the dataset used as depicted in the Figure 6: Prototype Model-2. 6. PROPOSED METHOD In this project, we propose to develop a system that could do ranking of the twitter users based on various algorithms taking different factor into consideration. Custom ranking algorithm is applied to generate ranking of most extremist users by taking the basis and cognizance from knowledge graph. In this we use some content based and source based
  • 4. features of tweets to rank the twitter users for their extrem- ism and impact. The content based features include the polarity of sentiment and a categorical feature which classi- fies any tweet as ’Normal’, ’Moderate’ or ’Extremist’. For this we use supervised algorithms where we trained on the manually annotated dataset. This helps in understanding the nature of the tweet. We assign labels as Label 0 for Neutral, Label 1 for Moderate, and Label 2 for Extreme. The source based feature include number of followers of the twitter user and the number of tweets the user made, which can provide insight on reachability of the user’s content. The degree of centrality and degree of betweeness are the two ma- jor parameters taken into consideration from the knowledge graph.In this project, we have used Twitter as the source of data collection, and this dataset is prepared by first scrap- ing the top trends from a website using beautiful soup after which theses trends are used to fetch tweets using tweepy library to create the entire dataset. It is very important to have a good representation and quality of data before applying the model for analysis. Thus Pre-processing is an important step that improves the generalizability of a model. 6.1 KNOWLEDGE GRAPH Knowledge graph contains set of vertices and edges where data points are interlinked and represents relation and pro- vides meaningful information. A graph G(V,E) where V rep- resents vertices i.e Twitter users, E represents the edges i.e represents the relation based on the mentions of the tweets by user. The degree of centrality and betweeness central- ity are the major parameters used. Betweeness centrality algorithm gauges the shortest path between each and every pair of nodes. The graph made is directed. This made use of (Breadth first search) BFS algorithm internally. Score is calculated for every score.This score is dependent on the shortest path that goes via the nodes. Higher betweeness centrality score is obtained when the nodes comes in be- tween the shortest path. Degree of betweeness is calculated using: Figure 3: Betweeness Centrality Centrality measures highly stressed numbers of paths.Centrality attempts to discover the most important nodes in a graph. centrality is calculated using: Figure 4: Centrality 6.2 PREPROCESSING • Cleaning of Tweets: Removal of URLs, Special Sym- bols, Links, Hashtags, Usernames. Twitter original data contains a lot of HTML entities thus, it is impor- tant to get rid of them. • Removal of Punctuation: Removal of unnecessary commas and symbols is very necessary in order to get cleaned Tweets. “.”, “,”,”?” are necessary punctuation and thus required to be retained while others have to be removed. • Removal of Duplicates: It is very necessary to re- move duplicate tweets as the data-set downloaded from Twitter contains the same retweets multiple times. It is a very important step to remove duplicate tweets from the Twitter data-set for further processing. • Word Tokenization: Before applying any of the mod- els, it is very important to tokenize the dataset so that we get the words that make processing easier by deep learning models. • Removal of Stop Words: Stop words are commonly used words such as a, an, in which occurs many times in a sentence, and such words do not add much mean- ing to sentences, so it is important to remove such words before applying deep learning algorithms in fur- ther stages. • Lemmatization: Morphological analysis is done by doing lemmatization by removing inflectional endings. This is done so that grouping of inflected items can be analyzed as a single item. For this, Vocabulary is used in a better way.
  • 5. Figure 5: Prototype Model - 1 Starting from duplicates 2467 tweets, we were left with 1068 distinct tweets after processing. These tweets were manually annotated into three categories namely Neutral, Moder- ate, and Extreme describing the level of extremism of the tweet.Then Word2Vec feature vector representation is ap- plied. In this project, we applied numerous Machine Learn- ing algorithms such as Multinomial Naive Bayes, Logistic Regression, Support Vector Machine, Decision Tree Clas- sifier, Random Forest to predict the level of extremism in any tweet. The training and testing accuracy are depicted in the below Result Table 1. Label prediction is made by the models, and later these predicted labels are used in the ranking algorithm. This feature is then used in the Figure 5: Prototype Model - 3 for generating the ranking of twitter users. 6.3 RANKING GENERATION For ranking custom based ranking algorithm is used as shown in figure 5: Prototype Model-2 It generates scores for each user available in the dataset, depening upon source-based features i.e., the user and content-based features i.e., the tweets by the user. The source based features consist of number of followers The con- tent based features include polarity of sentiment of tweet and the level of extremism that the tweet belongs to. The features are normalized so that one feature does not dominate over others. Knowledge graph is being constructed on those tweets dataset. It takes two major parameters into consideration i.e. degree of centrality and degree of betweeness. Another major factor taken into consideration is Tweet Count. Thereafter final score is generated using all these features and then the ranking is done on the ba- sis on the decreasing value of score as shown in Figure 5: Prototype Model - 2. Highest score implies top rank. The baseline results are shown in Figure 2. Figure 6: Prototype Model - 2 7. CONCLUSION The project helped in identifying the most extremist twitter users with the ranking generated. The ranking got improved with the cognizance and output from the Knowledge graph and given to the custom ranking algorithm. The increased accuracy and precision of ranking was due to the extra num- ber of parameters included in the algorithm. The idea of knowledge graph found to be very effective as it helped in identifying the importance of the node. Between centrality and degree centrality was major parameter in the analysis of knowledge graph.The deep learning architecture was proved to be better in analysis in comparison to the baselines mod- els. Class prediction using model MLP helped in identify- ing the categories which provided us the information about which tweets were moderate, neutral or highly extremist. The above algorithm is thus effective in doing analysis of threat level of twitter users during the high impact event. The UI shows the top ranked extremist twitter users on the basis of hashtags fetching and its tweets collected, by apply- ing the custom ranking algorithm. This indicates that these users are highly responsible for generating the threat level during elections and have high impact during these times. 8. RESULTS
  • 6. Figure 7: Ranked Result 9. EVALUATION The baseline results for five different models are shown below with their training and testing accuracy. Baseline Models Models Training Accuracy Testing Accuracy Naive Bayes 86.29 73.37 Support Vector Machine 86.29 75.92 Logistic Regression 98.04 77.34 Decision Tree Classifier 98.74 76.49 Random Forest 98.74 76.2 Table 1: Result Table Application of Deep Learning model Multi-Level perceptron with improved training and testing accuracy. Deep Learning Model Models Training Accuracy Testing Accuracy Multi-Level Perceptron 99.5 82.4 Table 2: New Result Table 10. FUTURE WORK • Knowledge graph provides a base which is used to strengthen its search results. Improve the speed of generating analysis from Knowledge graph,because de- creased speed may hamper user experience. As the knowledge graph is applied on big dataset so it con- sumes a lot amount of time there is need to improve and speed up so that it can be prevented from ham- pering the performance. • Add more analysis on tweets to produce better results. More analysis can be done by considering the retweet just like mentions. Further many different computa- tions can be done on Knowledge graph like Eigen vec- tor analysis. 11. REFERENCES [1] M. Cinelli, S. Cresci, A. Galeazzi, W. Quattrocioc- chi, and M. Tesconi. The limited reach of fake news on twitter during 2019 european elections. PloS one, 15(6):e0234689, 2020. [2] M. D. Conover, B. Goncalves, J. Ratkiewicz, A. Flam- mini, and F. Menczer. Predicting the Political Alignment of Twitter Users. In 2011 IEEE Third Int’l Conference on Privacy, Security, Risk and Trust and 2011 IEEE Third Int’l Conference on Social Computing, pages 192– 199, Boston, MA, USA, Oct. 2011. IEEE. [3] A. Gupta and P. Kumaraguru. Credibility ranking of tweets during high impact events. In Proceedings of the 1st Workshop on Privacy and Security in Online Social Media - PSOSM ’12, pages 2–8, Lyon, France, 2012. ACM Press. [4] M. Nouh, J. R. C. Nurse, and M. Goldsmith. Under- standing the Radical Mind: Identifying Signals to De- tect Extremist Content on Twitter. arXiv:1905.08067 [cs, stat], May 2019. arXiv: 1905.08067. [5] V. Pérez-Rosas, B. Kleinberg, A. Lefevre, and R. Mihalcea. Automatic Detection of Fake News. arXiv:1708.07104 [cs], Aug. 2017. [6] S. Rudinac, I. Gornishka, and M. Worring. Multimodal Classification of Violent Online Political Extremism Content with Graph Convolutional Networks. In Pro- ceedings of the on Thematic Workshops of ACM Multi- media 2017 - Thematic Workshops ’17, pages 245–252, Mountain View, California, USA, 2017. ACM Press.