IRJET - Unauthorized Terror Attack Tracking System using Web Usage Mining

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 07 Issue: 04 | Apr 2020 www.irjet.net p-ISSN: 2395-0072
© 2020, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 792
Unauthorized Terror Attack Tracking System using Web Usage Mining
Akash Yadav1, Minal Chudasama2, Dipti Patil3, Ranjita Gaonkar4
1-3BE Student, Department of Information Technology, MES Pillai College of Engineering, Navi Mumbai, India
4Assistant Professor, Department of Computer Engineering, MES Pillai College of Engineering, Navi Mumbai, India
---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract - With the speedy increase in social network
applications, people are using these platforms to voice their
choices regarding the daily problems. Gathering and
Analyzing people’s reactions towards shopping for a product,
public services, so on are very important. Sentiment Analysis
(or Opinion Mining) may be a common dialogue preparing
task that aims to discover the feelings behind opinions in texts
on varying subjects. In recent years, researchers within the
field of sentiment analysis are involved with analyzing
opinions on various topics like movies, commercial products
and daily social group problems. Twitter is a tremendously
popular micro blog on which users might voice their opinions.
Opinion investigation of twitter data is a field that has been
given a lot of attention over the last decade and involves
dissecting “tweets” (comments) and the content of those
expressions. Solely this, Paper explores the assorted sentiment
analysis applied to twitter data and their outcomes.
Key Words: TextBlob, Vader Sentiment, Natural
Language Processing, Sentiment, Analysis, Sentiment
Analysis, CSV, Tweets, Twitter, Countries, Terrorism,
Terror.
1. INTRODUCTION
Day to day the utilization of internet goes on increasing
drastically, comparatively the massive range of techniques
are there rising each day on varied dynamic platforms. Use
of Internet is expanding however, together with these
obstructive minded people are victimizing the net for
hurting to the society and people. It is believed that the
detection of terrorist’s activities on the Internet may stop
more terrorist attacks. Several terrorists or terrorist teams’
square measure createdtousesuchtechniques,applications,
and web to unfurl the fear. They traditionally attract the
youths to be entangled in such activities. It is necessary to
notice such attempts as to stop such hazardous things.
Terrorists utilize social sites. It's a massive challenge to
notice such venturesome terrorist attacks. It's similar to
eavesdropping.
Sentiment analysis is also called “opinion mining” or
“Emotion Artificial Intelligence”andalludestotheutilization
of Natural Language Processing (NLP), text mining,
computational linguistics, and bio measurements to
methodically acknowledge, extricate, evaluate, andexamine
emotional states and subjective information. Sentiment
analysis mostly involves the voice in client materials; for
example, surveys and reviews on the Web and web-based
social networks.
2. LITERATURE SURVEY
1. Naseema Begum A, Hanu Rakavi S, Mohanambal R,
Aswathy R, “Detection of online spread of terrorism using
web data mining”, [1]
The application that is to be developed will stop the spread
of terrorism by preventing the radicalization of the people
through online websites and the other media available on
the internet. People are denied access to these websites and
these media because the terrorist organizations are using
the internet as the media to promote and spread terrorism
and hence the terrorism is avoided. Data miningtechniqueis
utilized to determine and define the pattern from the
available collection of the websites and the data from the
websites mined are the huge volume of data sources, the
results obtained are widely used [1]. Web mining is also
similar to data mining since it involves the text mining
methods that are used to scan the data and also extracting
the useful pattern from unstructured data
2. Abdullah Alsaeedi, Mohammad Zubair Khan,”A Study on
Sentiment Analysis Techniques of Twitter Data” [2].
The diverse techniques for Twitter sentiment analysis
methods were discussed, including machine learning,
ensemble approaches and dictionary (lexicon) based
approaches. Also, hybrid and ensemble Twitter sentiment
analysis techniques were explored. Research outcomes
demonstrated that machine learning techniques; for
example, the SVM and MNB produced the greatestprecision,
especially when multiple features were included. SVM
classifiers may be viewed as standard learning strategies,
while dictionary (lexicon) based techniques are extremely
viable at times, requiring little efforts in the human-marked
archive. Machine learning algorithms, such as The Naive
Bayes, Maximum Entropy, and SVM, achieved an accuracy of
approximately 80% when n-gram and bigram models were
utilized. Ensemble and hybrid-based Twitter sentiment
analysis algorithms tended to perform better than
supervised machine learning techniques, as they were able
to achieve classification accuracy approximately.
3. T. Suvarna Kumari, Narsaiah Putta, “Web Data Mining to
Detect Online Spread of Terrorism” [3].
This paper talks about a method for mental oppressor area
by using Web traffic substance and using data mining
techniques because the survey information is seenhere. The
suggested approach, called ATDS, knowspsychicoppressors

standard result (profile) by adding a knowledge mining
estimate to the writing of Web objectives related to dread.
The resulting picture including records is employed by the
framework to execute ongoing customerrevelationlinkedto
being crowded with functions out of fear monger.
Examination of the Receiver Operator Characteristic (ROC)
shows that this scheme can defeat a framework of intrusion
recognition centered on going. Thusly, fear related activities
are often perceived using data mining frameworks and log
access and handling.
4. Harshali P. Patil, Dr. Mohammad Atique “Sentiment
Analysis for Social Media: A Survey”, [4].
Masses of users share their feelings on social media, making
it a valuable platform for trailing and exploring public
sentiment. Social media is one in all the largest platforms
where massive instant messages are revealed everyday
which makes it a perfect supply for capturing the opinions
towards varied curious topics, like product, goods or
celebrities, etc. the most goal of this paper is to administer
associate overview of latest updates in sentiment analysis
and classification strategies and it includes the temporary
and brief discussions on the challenges ofsentimentanalysis
that the work wants to be done. We tend to additionally find
that the majority of the works done area unit based on
machine learning methodology instead of the lexicon based
methodology. Here, detailed survey on different techniques
used in Sentiment Analysis is carried out to understand the
level of work.
5. Ramesh Yevale, "Unauthorized Terror Attack Tracking
Using Web Usage Mining", [6].
The purpose of the project is that the project is incredibly
helpful to search out the terror related activities. By this
project, users of the project will come across to understand
the suspicious spoken conversation, and therefore will
notice the supply of the terror and can be helpful to
attenuate the fear activities. The project has terribly wide
scope within the security of the national assets. The project
is incredibly helpful to search out terroractivities.It’schiefly
helpful for the Military and CBI agents. By mistreatment
they'll come back to understand the various terror
communication and therefore the accidents will be avoided.
Blogs, social networking sites, largely hosted by terrorist
sympathizers is avoided by mistreatment net usage mining.
3. SYSTEM DESIGN
The website that we built starts with a home page that
highlights the key features ofthewebsite,leadinguserstothe
other pages which provide them with different features. The
three main features ofourwebsitearenamely,GlobalSearch,
Country Search and Upload Dataset. These servicesthenlead
the user to the further functioning of the system.
Fig 1:- Proposed System
3.1 Sentiment Analysis
Sentiment analysis [7] is to analyze the collection of these
personal preference expressions, also known as emotion
mining and opinion extraction. According to the size of the
text, the sentiment analysis can be divided into three
categories: 1) sentiment analysis based on the words. It
creates a sentiment lexicon to separate positive words and
negative words by hand or automatically, and then uses the
lexicon to analyze the data based on the frequency of the
word or the sentiment value which is set when building the
lexicon. 2) sentiment analysis based on the sentences. It
mainly analyzes the relevant factors of the sentiment
tendency such as the semantic expression and sentence
representation of sentences and so on, combined with the
polarity of the modifier and the negative words.3)sentiment
analysis based on the document. Basically, it’s similar to the
second category, because the statement can be seen as short
documents. In this paper, sentiment analysis involves
different techniques such as the establishment of sentiment
lexicon, the sentiment analysis based on the words and the
label of the sentiment value such as positive, negative or

neutral. The sentiment analysis of the word is the basisofthe
sentiment analysis of the text, and the premise of sentence
and chapter emotional analysis.
3.2 TextBlob in Python
TextBlob is a python library and offers an easy API to
access its methods and perform basic Natural Language
Processing tasks. TextBlob is a Python (2 and 3) library for
processing textual data. It provides an easy API for dividing
NLP tasks like part-of-speech tagging, phrase extraction,
sentiment analysis, classification, translation, and more.
TextBlob and Naive Bayes Classifier areusedforthisprocess.
3.3 Naïve Bayes Classifier
Bayesian classification provides practical learning
algorithms and previous knowledge and observed data are
often combined. Bayesian Classification provides a helpful
perspective for understanding and evaluating several
learning algorithms. It calculates explicit possibilities for
hypothesis and it's sturdy to noise in input file.
1. Consider a training data setD consists of documentswhich
belongs to different classes say class A and B.
2. Prior probability of both classes A and B is calculated as
shown Class A=number of objects ofclass A/ totalnumberof
objects. Class B=number of objects of class B / total number
of objects.
3. Now calculate the totalnumberofwordfrequenciesofboth
classes A and B i.e., ni na = the total number of word
frequency of class A. nb =the total number of wordfrequency
of class B.
4. Calculate the conditional probability of keyword
occurrence for given class P(word1 / class A) = word count /
ni(A) P(word1 / class B) = word count / ni(B) P(word2 /
class A) = word count / ni(A)P(word2/classB)=wordcount
/ ni P(wordn / class B) = wordcount / ni (B)
5. Uniform distributions are to be performed in order to
avoid zero frequency problem.
6. Now a new document M is classified based on calculating
the probability for both classes A and B P (M/W). a) Find P(A
/ W) = P(A) * P(word1/class A)* P(word2/ class A)……*
P(wordn / class A). b) Find P(B / W) = P(B) * P(word1/class
B)*P(word2/ class B)……* P(wordn / class B).
7. After calculating probability for both classes A and B the
class with higher probability is the one the new document M
assigned.
3.4 VADER Sentiment
VADER (Valence Aware Dictionary and Sentiment
Reasoner) is a lexicon and rule-based sentimentanalysistool
that is specifically attuned to sentiments expressed in social
media. It is absolutely open-sourced. VADER uses a
combination of a sentiment lexicon is a list of lexical options
(e.g., words) which are generally labeled according to their
semantic orientation as either positive or negative. VADER
not solely tells regarding the Positivity and Negativity score
however additionally tells about how positive or negative a
sentiment is Manually making (much less, validating) a
comprehensive sentiment lexicon is a labor intensive and
generally error prone method, therefore there is no surprise
that several opinion mining researchers and practitioners
rely heavily on existing lexicons as primary resources.
The Python code for the rule-based sentiment analysis
engine. Implements the grammatical and syntactical rules
represented within the paper, incorporating empirically
derived quantifications for the impact of every rule on the
perceived intensity of sentiment in sentence-level text.
Significantly, these heuristics transcend what would usually
be captured in an exceedingly typical bag-of-words model.
They incorporate word-order sensitive relationships
between terms. For instance, degree modifiers (alsoreferred
to as intensifiers, booster words, or degree adverbs) impact
sentiment intensity by either increasing or decreasing the
intensity.
The algorithm goes as such:
1. Count the words in the sentence.
2. Generate the score for the all the words in the sentence
individually.
3. Calculate the sum of all the word scores.
4. Generate a compound score which shows if the sentenceis
positive / negative / neutral.
The compound score is computed by summing the
valence score of every word within the lexicon, adjusted
consistent the rules and so normalized to be between -1
(most extreme negative) and +1 (most extreme positive).
This is the foremost helpful metric if you wish one one-
dimensional live of sentiment for a givensentence.Callingita
'normalized, weighted composite score' is correct.
It is also useful for researchers who would like to set
standardized thresholds for classifyingsentencesaspositive,
neutral, or negative. Typical threshold values (used in the
literature cited on this page) are:
1.Positive sentiment: compound score >= 0.05
2.Neutral sentiment: (compound score > -0.05) and
(compound score < 0.05)
3. Negative sentiment: compound score <= -0.05
The pos, neu, and neg scores are ratios for proportions of
text that fall in eachcategory (so these should all adduptobe
1... or close to it with float operation). These are the most
useful metrics if you want multidimensional measures of
sentiment for a given sentence
4. IMPLEMENTATION DETAILS
4.1 GUI
This is the overall UI of the application. The website is
made by using HTML, CSS and CSS libraries like Bootstrap,
Flask and JavaScript. Users will interact with this part of the
application only, which is the website. On the HomePage,the
user has the option to pick one of 3 given ways to provide
data, The Global Search, Country Search and the Upload
Dataset.

4.2 Global Search
Global Search is the first option to download Tweets from
Twitter. The user downloadsthetweetbyenteringthesearch
Keyword in the form provided.Thisquerywillthenbepassed
to the Python function called download tweets. This function
will use the search Keyword providedtodownloaddatafrom
Twitter, withoutRetweets.Afterthetweetsaredownloadeda
CSV file will be created. This CSV file will be preprocessed
again and a final CSV file will be created which will be later
displayed on the preprocessed table page.
After the CSV file is created download tweets () function
will redirect the user to the table page which displays the
preprocessed data in a tabular form andalinkintheformofa
button to the next page. When the user clicks on the Evaluate
Sentiment button it runs the tweets graph() function which
performs sentiment analysis on the preprocessed data using
both Vader Sentiment and Text Blob techniques and then
writes the twitter username, tweet and the sentiment values
for each tweet to the CSV files and redirects the user to the
next page.
The next page consists of a bar graph generated from the
sentiment values using JavaScript and the following
JavaScript libraries: c3.js, papa parse and d3.js, a table and a
pie chart. The bar graph and the table is displayed on the
page with an option to download the table. A pie chart is also
generated from the statistics of tweets downloaded and the
statsare displayed alongsidethe bar graphandthetable.The
user can also view a summary of the tweets from 5 preset
major countries relevant to the search query providedbythe
user. This summary page has a bar graph which shows
number of positive, neutral and negative tweets for each
country and a pie chart with statistics and a summary table.
4.3 Country Search
The country search feature is similar to the global search
feature but instead provides tweets from 5 countries and
displayed one at a time and the user can selectwhichcountry
tweets to view. When the user enters the search query the
query is passed to the python function country tweets() and
this function downloads the tweetsfor5presetcountriesand
saves the tweets to the CSV files for each country and then
another function country sentiment() runs which performs
sentiment analysis on the preprocessed data using both
Vader Sentiment and Text Blob techniques and then writes
the twitter username, tweet and the sentiment values for
each tweet to the CSV files and redirects the user to the next
page.
The next page consists of a bar graph generated from the
sentiment values, a table and a pie chart. The bar graph and
the table is displayed on the page with anoptiontodownload
the table. A pie chart is also generated from the statistics of
tweets downloaded and the stats are displayedalongsidethe
bar graph and the table. The user can choose the country
from which they want to view the data from the drop down
menu.
4.4 Upload Dataset
The upload dataset feature allows the user to provide
their own dataset and it is preprocessed for the sentiment
visualizer which displays the sentiments for the uploaded
dataset. The user can upload their dataset in a CSV format
using the upload form and when the user hits upload the
python function processing() runs and the CSV file is
preprocessed and the required fields are extracted from the
provided dataset and placed in the appropriate fields in the
preprocessed CSV file. Then sentiment analysis is performed
on the preprocessed data and the sentiment values are
written to a new CSV file. The user is then redirected to the
next page.
The next page consists of a bar graph which shows the
sentiment of each tweet alongside the table and the pie chart
which also consists of the statistics showing the number and
percentage of positive, neutral and negative tweets in the
provided dataset. The user can also download the CSV file
from this page.
Fig 2:- System Functional Module
5. PROJECT INPUT AND OUTPUT SCREENSHOTS
5.1 Home Page
Our websiteprovides theservicestoperformSentimental
Analysis on user generated data. The process of sentimental
analysis helps us define the general idea of the nature of the
statement on social media in our case tweets. There are 3
services to do, Global search, Country search and Update
Dataset.
Fig 3:- Home Page (1)

5.2 Global Search
When enacted on this Global Search page by the user; this
module allows the user to enter search any term, related to
which the user wishes to obtain data or in our case Tweets.
The user enters the search query, which results in a hash tag
search on twitter using the Python library “tweepy”.
Fig 6:- Global Search Page
5.3 Country Search
This module provides user with the option to download
tweets related toany term but for5specificpredefinedmajor
countries. This feature is very useful for research purposes
and makes the data more refined and country specific. The
user can select to view one country at a time from the
dropdown menu.
Fig 7:- Country Search
5.4 Upload Dataset
In Upload Dataset, there is an option for the user to
providethe data in the form of CSV file. Thefurtherprocessis
then done similarly, on the provided data CSVfilebytheuser.
Fig 8:- Upload Dataset
5.5 Preprocessed Data Table
The resultanttweetsaregatheredanddownloadedoneby
one into a CSV file in a raw format. This CSV file isa collection
of preprocessed data. The output CSV file consists of two
fields: the twitter username and the corresponding tweet.
Only the required fields are selected ignoring the duplicate
data and this new set of data is then written into another CSV
file. This creates our preprocessed data. From this data we
can perform Sentimental Analysis.
Fig 9:- Preprocessed Data Table
5.6 Graph Generation
This process altogether creates a clean version of the
sentiments generated from the content of the tweets making
it visual friendly by creating a graphical format of the two
algorithms used namely, TextBlob and VADER sentiment

which displays the data over the course of time. The graph
displays, The Positive sentiment,whicharethetweetshaving
sentimental value 15, the Negative sentiment, which are
tweets with tweets with value 5 and Neutral sentiment,
tweets having sentimental value 10 and a pie chart which
displays all the distribution of positive, neutral and negative
sentiment percentage in the dataset. Similarly, we can view
the result for Country specific search.
Fig 10:- Results – Graph
Fig 11:- Results – Pie chart
Fig 12:- Results- Graph (Country Search)
5.7 Download Result Table
The expected result generated is also represented in a
tabular form for better understanding and convenience and
can be downloaded for further use by the user. The table
contains fields like Username, Tweet and the resultant
sentiment.
Fig 13:- Result Table
Fig 14:- Downloaded Result Table
5.8 Countries Overview
The result could also be further divided in to country
specific data and graph as shown in the figure 10, Countries
Overview. When this function runs it downloads tweets for
the search term provided by the user for 5 different
countries. The data is distributed into 5 pre-defined major
countries.
Fig 15:- Countries Overview
6. CONCLUSIONS
In conclusion, we have developed an application that
performs Sentiment analysis on tweets posted by the users
round the world and conjointly in specific countries and so
this displays the tabular information within the style of
interactive graphs. The sentiments are calculated using
Natural Language Processing (NLP) techniques such as
TextBlob and Vader sentiment that are very accurate. We
need to examine the assorted Data Mining and Web Mining
technologies and see how they can be adapted for counter-
terrorism, for future scope.

ACKNOWLEDGEMENT
We would like to thank our mentor, Prof.Ranjita Gaonkarfor
guidance and unwavering support throughout the project
and the semester. We would liketothank ourHOD,Dr.Satish
Kumar L. Varma for their encouragement and motivation to
learn and implement projects of sorts. Lastly, we would like
to thank our principal, Dr. Sandeep Joshi for providing us
opportunities to explore our domainandformotivatingusto
do better.
REFERENCES
[1] Naseema Begum A, Hanu Rakavi S, Mohanambal R,
Aswathy R, “Detection of online spread of terrorism
using web data mining”, IEEE 2019.
[2] Abdullah Alsaeedi, Mohammad ZubairKhan,”AStudyon
Sentiment Analysis Techniques of Twitter Data”, IEEE
2019.
[3] T. Suvarna Kumari, Narsaiah Putta, “WebData Miningto
Detect Online Spread of Terrorism”, IEEE July 2019.
[4] Sonali Vighne, Priyanka Trimbake, Anjali Musmade,
Ashwini Merukar, Sandip Pandit, "An Approach to
Detect Terror Related Activities on Net", IEEE 2016.
[5] Harshali P. Patil, Dr. Mohammad Atique “Sentiment
Analysis for Social Media: A Survey”, IEEE 2015.
[6] Ramesh Yevale, "Unauthorized Terror Attack Tracking
Using Web Usage Mining", IEEE 2014.
[7] Sentiment analysis, In Wikipedia. Retrieved November
2, 2017,from
https://en.wikipedia.org/wiki/Sentiment_analysis

IRJET - Unauthorized Terror Attack Tracking System using Web Usage Mining

More Related Content

What's hot (20)

Similar to IRJET - Unauthorized Terror Attack Tracking System using Web Usage Mining (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET - Unauthorized Terror Attack Tracking System using Web Usage Mining