INTERNATIONAL RESEARCH JOURNAL OF ENGINEERING AND TECHNOLOGY (IRJET) E-ISSN: 2395-0056

VOLUME: 07 ISSUE: 12 | DEC WWW.IRJET.NET P-ISSN: 2395-0072

2020

Automated Detection of Cyberbullying Using Machine Learning

Niraj Nirmal1, Pranil Sable2, Prathamesh Patil3, Prof. Satish Kuchiwale4
1-4SIGCE Maharashtra, INDIA.
------------------------------------------------------------------------***-----------------------------------------------------------------------
Abstract— Increasing the use of Internet and facilitating where people engage in social interaction, offering the
access to online communities such as social media have led to possibility to establish new relationships and maintain
the emergence of cybercrime. Cyberbullying is very common existing friendships. On the negative side however,
now a days. which have no tracking like it may harm any social media increase the risk of children being
individual, business, society, country in past few days it seems confronted with
that riots were happened due to some statement used by one
community on another its important to identify such content
which spreads hate or harm community text processing, NLP
(natural language processing) is an emerging field with the
help of NLP and machine learning algorithms such as naive
bayes, random forest, SVM we are going to identify
cyberbullying in twitter. Objectives of this implementation
written in objective section. Image character with the help of
OCR will be done by us to find image - based cyberbullying
the impact on individual basis thus will be checked on dummy
system. Machine learning and natural language processing
techniques to identify the characteristics of a cyberbullying
exchange and automatically detect cyberbullying by matching
textual data to the identified traits. On the basis of our
extensive literature review, we categorise existing approaches
into 4 main classes, namely supervised learning, lexicon-
based, rule-based, and mixed-initiative approaches.
Supervised learning-based approaches typically use
classifiers such as SVM and Naïve Bayes to develop
predictive models for cyberbullying detection.

Index Terms— cyber bullying, natural language

processing, machine learning algorithms, Social
networking.

I. Introduction

It is not sufficient to remind students of regulations

forbid- ding plagiarism; In recent years, the use of
social networking increased. And social networking
sites are great tools of connecting to people.
However, as social networking has become
widespread. People are finding illegal and
unethical ways to use these communities. We see
that people, especially teens and young adults, are
finding new ways to bully one another over the
Internet. Bullying is not a new phenomenon and
cyber bullying has manifested itself as soon as
digital technologies have become primary
communication tools. On the positive side, social
media like blogs, social networking sites (e.g.
Facebook), and instant messaging platforms (e.g.
WhatsApp) make it possible to communicate with
anyone and at any time. Moreover, they are a place

© 2020, | Impact Factor value: | ISO 9001:2008 Certified | Page 1

 INTERNATIONAL RESEARCH JOURNAL OF ENGINEERING AND TECHNOLOGY (IRJET) E-ISSN: 2395-0056

VOLUME: 07 ISSUE: 12 | DEC WWW.IRJET.NET P-ISSN: 2395-0072

2020
threatening situations including grooming or bullying without understanding the larger context
sexually transgressive behaviour, signals of of the exchange, even though the message is
depression and suicidal thoughts, and clearly expressing very negative sentiments.
cyberbullying. Users are reachable 24/7 and Conversely, positively-expressed sarcasm.
are often able to remain anonymous if desired:
this makes social media a convenient way for
bullies to target their victims outside the
school yard. The detection of cyberbullying
and online harassment is often formulated as
a classification problem. Techniques typically
used for document classification, topic
detection, and sentiment analysis can be used
to detect electronic bullying using
characteristics of messages, senders, and the
recipients. It should, however, be noted that
cyberbullying detection is intrinsically more
difficult than just detecting abusive content.
Additional context may be required to prove
that an individual abusive message is part of a
sequence of online harassment directed at a
user for such a message to be labelled as
cyberbullying. The growth of cyberbullying
activities is increasing as equally as the
growth of social networks. Cyberbullying
activities poses a significant threat to mental
and physical health of the victims. Project
about detection of bullying is present but
implementation for monitoring social network
to detect cyberbullying activities is less.
Hence, the proposed system focuses on
detecting the presence of cyberbullying
activity in social networks using natural
language processing and machine learning
algorithms which helps government to take
action before many users becoming a victim of
cyberbullying. Detection of cyberbullying and
the provision of subsequent preventive
measures are the main courses of action to
combat cyberbullying. The proposed method is
an effective method to detect cyberbullying
activities on social media. The detection
method can identify the presence of
cyberbullying terms and classify cyberbullying
activities in social network such as Flaming,
Harassment, Racism and Terrorism using
natural language processing and machine
learning algorithms. Cyberbullying detection
is inherently difficult due to the subjective
nature of bullying. It extends beyond detecting
negative sentiments or abusive content in a
message as these tasks, on their own, do not
necessarily mean that the message is in fact
bullying. For example, a message such as “I’m
disgusted by what you said today and I never
want to see you again” is difficult to classify as

© 2020, | Impact Factor value: | ISO 9001:2008 Certified | Page 2

 INTERNATIONAL RESEARCH JOURNAL OF ENGINEERING AND TECHNOLOGY (IRJET) E-ISSN: 2395-0056

VOLUME: 07 ISSUE: 12 | DEC WWW.IRJET.NET P-ISSN: 2395-0072

2020

II. Related Work

Table: Literature Survey

Title Authors Problem Solution Result

An Effective Approach Divyashree, Vinutha H, The biggest problem In this paper focused on represented a novel
for Cyberbullying Deepashree N S regarding cyberbullying is the issues of robust method on the current
Detection and that the age group of the system and objectives are scenario of cyber-
avoidance offenders ranges from as 1) Automatic detection bullying and various
young as eight to the legal and avoidance of methods available for the
adult age of eighteen and cyberbully attack in detection and prevention
beyond. Once happen this internet. of cyber harassment. Our
activity then victims are 2) Effective age concept depends upon
often left permanently authentication for the text analysis, the data
then difficult to find them. website browsing and which is uploaded or text
categorizing the links written by any user is
based on age. 3) Effective first analyzed.
website filtering in
search results based on
ranking. 4) Enhanced

searching
procedure

promisingly reduces the

effort of user in
searching indented
websites.
Using Machine Kelly Reynolds, April teens and young adults, Used machine learning used a language-based
Learning to Detect Kontostathis, Lynne are finding new ways to algorithm to method of detecting
Cyberbullying Edwards bully one another over the detect cyberbullying. By
Internet. in a study cyberbullying. For recording the percentage
conducted by Symantec training the data of curse and insult words
reported that, to their downloaded from within a post.
knowledge, their child has website. The data was
been involved in a labeled using a web
cyberbullying incident. service. the labeled data,
in conjunction with
machine learning
techniques provided by
the Weka tool kit, to train
a computer to recognize
bullying content.
Cyberbullying Detection Liew Choong Hon, Kasturi Increased cyberbullying this system, the users with the advent of this
System on Twitter Dewi Varathan attacks on the social can identify the cyberbullying detection
network services. To cyberbullying related and solution system in
prevent these activities tweets based on the Twitter, it will help the
proposed an system. keywords and populate it authorities to monitor,
in a news feed form. By regulate or at least
doing this, it allows users decrease the harassing
to determine the incidents in cyberspace
identities of the
cyberbullies and the
victims
from the cyberbullying
tweets
Automatic detection of Cynthia Van Hee,Gilles Increased the cyberbullying The focus of this paper is In this paper investigate
cyberbullying in social Jacobs,Chris Emmery,Bart using on automatic the automatic detection
media text Desmet, Social media sides/apps. cyberbullying of cyberbullying-related
detection in social media posts on social media.
text by modelling posts Given the information
written by bullies, overload on the web,
victims, manual monitoring for
and cyberbullying has
bystanders of online become unfeasible.
bullying. In this paper
support vector machine Automatic detection of

© 2020, | Impact Factor value: | ISO 9001:2008 Certified | Page 3

 INTERNATIONAL RESEARCH JOURNAL OF ENGINEERING AND TECHNOLOGY (IRJET) E-ISSN: 2395-0056

VOLUME: 07 ISSUE: 12 | DEC WWW.IRJET.NET P-ISSN: 2395-0072

2020
is used to exploiting a signals of cyberbullying
rich feature set and would enhance
investigate which moderation and allow to
information sources respond quickly when
contribute the most for necessary.
the task.
Methods for detection of Rekha Sugandhi, Anurag major problem when it This paper aims to review In this paper realize
cyberbullying: A survey Pande, Siddhant Chawla, comes to cyber bullying is the different methods support vector machines
Abhishek Agrawal, Husen the lack of identifiable and algorithms used for have given the best
Bhagat parameters which mark detection in cyber result. We plan to
any post as a bullying bullying and provide a implement SVM in our
instance. comparative study project as the primary
amongst them so as to classifier for our base
decide which method is dataset.
the most effective
approach and provides
the best accuracy.

© 2020, | Impact Factor value: | ISO 9001:2008 Certified | Page 4

 INTERNATIONAL RESEARCH JOURNAL OF ENGINEERING AND TECHNOLOGY (IRJET) E-ISSN: 2395-0056

VOLUME: 07 ISSUE: 12 | DEC WWW.IRJET.NET P-ISSN: 2395-0072

2020

2.1 Aim of the Project studies report that cyberbullying constitutes a growing
problem among youngsters. Successful prevention
The main aim of the detecting the cyberbullying depends on the adequate detection of potentially
model will help to improve manual monitoring for harmful messages and the information overload on the
cyberbullying on social networks. In this project we Web requires intelligent systems to identify potential
fetch the tweets from twitter accounts and risks automatically. So, In this project we focus on to
preprocess the twits and images and applying make a model on automatic cyberbullying detection in
generated model will detect the cyberbullying or social media text by modelling posts written by bullies
not. on social network.
The objectives of the systems development and
event management are:

Collect the dataset of bullying words and

preprocess it and apply natural language
processing and then machine learning algorithms
Generate different machine learning algorithm
model.

Fetch the tweets from twitter account and

preprocess it.

Apply generated model on the fetched tweets and

get final output cyberbullying or not.

2.2 Scope of the Project

Cyberbullying is the use of electronic

communication to bully a person by sending
harmful messages using social media, instant
messaging or through digital messages.
Cyberbullying can be very damaging to adolescents
and teens. It can lead to anxiety, depression, and
even suicide. Also, once things are circulated on
the Internet, they may never disappear,
resurfacing at later times to renew the pain of
cyberbullying. Cyberbullying can be very damaging
to adolescents and teens. It can lead to anxiety,
depression, and even suicide. Also, once things are
circulated on the Internet, they may never
disappear, resurfacing at later times to renew the
pain of cyberbullying. So overcome these issues
detecting the cyberbullying is very important in
now a days which will help to stop cyberbullying on
social media networks.

2.3 Problem Statement

The social media network gives us to great

communication platform opportunities they also
increase the vulnerability of young people to
threatening situations online. Cyberbullying on an
social media network is a globle phenomenon
because of its huge volumes of active users. The
trend shows that the cyber bullying on social
network is growing rapidly every day. Recent

© 2020, | Impact Factor value: | ISO 9001:2008 Certified | Page 5

 INTERNATIONAL RESEARCH JOURNAL OF ENGINEERING AND TECHNOLOGY (IRJET) E-ISSN: 2395-0056

VOLUME: 07 ISSUE: 12 | DEC WWW.IRJET.NET P-ISSN: 2395-0072

2020
III. Methodology

This project we will develop using python and

web technology. Within that first we will
search and find the the dataset and download
it for train the model. After downloading first
we will pre-process the data and then
transferred to Tf-Idf. Then with the help of
naïve bayes, SVM (Support vector machine)
and DNN algorithm we train the dataset and
generate model separately. Then we are going
to develop a web based application using
FLASK framework. We will fetch the real time
tweets from twitter and then we apply
generated model to these fetched tweets and
check the text or images are cyberbullying or
not. These all-purpose we are using python as
backend, Mysql is database and for frontend
html, css, javascript etc.

Figure 1: Flow diagram of Cyberbullying

Detection

3.1 Technique of detection

3.1.1 Textual Based:

We group features such as cyberbullying

keywords, pro - fanity, pronouns, n-grams,
Bags-of-words (BoW), Term Frequency Inverse
Document Frequency (TFIDF), document
length, and spelling content-based features.

Content-based features are overwhelmingly

used across our sample, with as many as 41
papers utilising con-tent-based features. As
cyberbullying messages are often abusive and
insulting in nature, it is not surprising that
profanity was

© 2020, | Impact Factor value: | ISO 9001:2008 Certified | Page 6

 INTERNATIONAL RESEARCH JOURNAL OF ENGINEERING AND TECHNOLOGY (IRJET) E-ISSN: 2395-0056

VOLUME: 07 ISSUE: 12 | DEC WWW.IRJET.NET P-ISSN: 2395-0072

2020

found to be the most used content-based feature Dadvar and De Jong (2012), Sood and Churchill
across the reviewed studies,with 22 papers using (2012a), and Nahar et al.(2013).
the presence of profanity in text as an indicator for
cyberbullying. Studies such as Dinakar et al. Of the 41 studies using content-based features,
(2011), Perez et al.(2012), Kontostathis et al. 5checked for the presence of cyberbullying keywords as
(2013), Nahar et al.(2013)and Bretschnei-der et al. part of the detection process. By cyberbullying
(2014), created profanity lexicons using wordlists keywords, we refer to non-profane words the use of
compiled by the researchers or sourced from which can indicate the presence of cyberbullying. These
external libraries such as noswearing.com3 and often are words associated with themes such as race,
urban dictionary.com.By equating the presence of physical appearance, gender, and sexuality. As far back
profanity to cyberbullying, the use of profanity as the earliest study we discovered (i.e.,Mahmud et al.,
lexicon salone fails to consider other key aspects of 2008), cyberbullying key-words have
cyberbullying such as repeti- tiveness and the
presence of a power differential. Rafiq et al.
(2015)similarly cautioned against the use of
profanity as the only feature for cyberbullying
detection and argued that not all use of profanity
and cyber-aggression constitutes bullying. Studies
such as Nahar et al.(2013), Dadvar et al.(2014),
Bretschneider et al.(2014) and Nahar et al.
(2013)incorporate do ther features such as pro
nouns in close proximity to profanity, since such
personalised abusive content is potentially more
indicative of cyberbullying than abusive terms on
their own. For example, the phase “the f**king
train was delayed again” is definitely not
cyberbullying although it contained profanity but
“you f**king idiot” could be. While this is an
improvement, the pronoun + profanity feature still
suffers the same short com- ings as using profane
terms alone.

Dinakar et al.(2011),often cited for the

performance gain achieved by their label-specific
binary classifiers over multi- class classifiers,
achieved this improved performance by using
domain-specific content features learned from
training classifiers on a set of messages clustered
on sensitive topics such as race, culture, sexuality,
and intelligence to then detect bullying messages
within each cluster.

While Yin et al.(2009) did not find n-grams very

effective in their experiments, its use as a
detection feature is still relatively popular amongst
studies, including Dinakar et al. (2011), Xu et al.
(2012a; b), Sood and Churchill (2012a; b),and
Munezero et al. (2014).As TFIDF provides a
measure of a word’s importance to a document
within a collection of documents, it can sometimes
provide better results than using n-grams in
isolation (Yin et al., 2009). It is, therefore
,often used alongside n-gram and other features to
im-prove detection performance ,as can be seen in
the works of Yin et al.(2009), Dinakar et al. (2011),

© 2020, | Impact Factor value: | ISO 9001:2008 Certified | Page 7

 INTERNATIONAL RESEARCH JOURNAL OF ENGINEERING AND TECHNOLOGY (IRJET) E-ISSN: 2395-0056

VOLUME: 07 ISSUE: 12 | DEC WWW.IRJET.NET P-ISSN: 2395-0072

2020
been used as detection features and this trend different features. All features are assumed
has continued with later studies such as independent given label Y:
Dinakar et al. (2011), Sanchez and Kumar
(2011), Perez et al.(2012) and Dadvaret al. P(XI, ... , Xn/Y) = ∏ (Xi/Y)
(2013b).These studies created lexicons com-
posed of words so selected because their A very simple document representation is used
presence within a message or a post here, usually bag of words. Words important to the
connotesa high likelihood of cyberbullying. meaning of the text, and thus imperative in its
For example, both Dinakaret al. classification, are considered, and given weight
(2011)identified themes such as race, culture, according to meaning, or in this case, severity. For
sexuality, physical appearance ,and instance, ''faggof” would receive a higher weight
intelligence as common bullying topics and than "bitch", due to the former being sexually
used a lexicon of words associated with these discriminatory and abusive.
themes as features ,while Sanchez and Kumar
(2011) concentrated on homophobic slurs such
as “gay”, “queer”, “homo” and “dyke” as
keywords.[15]

3.1.2 Non textual Based:

While the focus of the studies in our sample

has largely been on textual bullying, images
and videos can also be used as delivery
systems for online bullying and their im-pact
can be as,or perhaps even more, damaging. In
addition, as social media platforms improve
their ability to detect and prevent textual
bullying, bullies may likely resort to the use of
other media forms to bypass anti-bullying
measures. Recent advances in image
processing and OCR (Optical Character
Recognition)make it viable to attempt
cyberbullying detection within media forms
like images, animations ,and videos .With
social media trends such as internet memes
and viral videos becoming hugely popular in
recent times, the scan be easily perverted by
bullies to perpetrate cyberbullying. We,
therefore ,envisage that developing systems
capable of detecting bullying content within
multimedia files is a key area for future
research considerations. [15]

IV. Approaches of Models

4.1. Naive Bayes Model

4.2. SVM Model

4.3. DNN Model

4.1 Naive Bayes Model:

The Naive Bayes family of classifiers are

simple conditional probabilistic classifiers that
work by applying Bayes theorem with naive
independence assumptions between the
© 2020, | Impact Factor value: | ISO 9001:2008 Certified | Page 8
 INTERNATIONAL RESEARCH JOURNAL OF ENGINEERING AND TECHNOLOGY (IRJET) E-ISSN: 2395-0056

VOLUME: 07 ISSUE: 12 | DEC WWW.IRJET.NET P-ISSN: 2395-0072

2020

Thus, given a document 'd' and class 'c':

•Linear kernel

P(c/d)= •Gaussian kernel

The maximum posterior class, or the most likely Linear kernel is a special case of the RBF kernel,
and works best when the number of features is
class, being in our case either bullying or not,
very large. The linear kernel on data sets acquired
would be:
from Myspace, Kongregate and Slashdot datasets
Cmap = were used. The datasets are available from the
workshop on Content Analysis for the Web 2.0 .
The datasets contain manually-labeled data from ,
= which is used as a ground truth dataset. Data from
3 different social networking sites are included in
the dataset: Slashdot (496
= files, 140,000 comments total (one for each article)),
Kongregate (12 files, 150,000 comments total (one for
each
The corpus of data obtained to experiment with is testing purposes is also converted into data matrix and
the same as that used for J48. In this case, a true this data matrix is passed to the classifier. SVMs use
positive rate of 0.723, taking into account both sophisticated statistical learning theory to overcome the
textual and social features, was obtained. Without curse of dimensionality
taking into account social features, the rate was
0.584 once again proving, as with similar tests Instead of specifying the feature vector, kernel
performed with J48, that social features help functions can be used to provide similarity between
improve the result.[18] data points. There are various kernels that can be used
with SVM namely,
4.2 SVM Model:
•RBF kernel (Radial basis function)
SVM (Support Vector machine) is a supervised
learning algorithm, and is one of the most efficient
and universal classification algorithms. Its goal is
to fmd the optimal separating hyperplane which
maximizes the margin of training data. Initially the
classifier is trained with labelled data before being
used to classify the data to test accuracy. Before
the data can be used to train our classifier, it is
imperative to process it. This consists of the
following steps:

•Labelling of data

•Generation of vocabulary

•Creation of document-term matrix

Once the labelled data is converted into a data

matrix based on the values in the vocabulary, the
values are then plotted and optimal hyperplane is
chosen based on the convex hull. The optimal
hyperplane is chosen in such a way that it
maximizes the margin of the training data. Once
the classifier is trained the input data is passed to
this classifier to segregate it into positive and
negative instances of bullying. This input data for

© 2020, | Impact Factor value: | ISO 9001:2008 Certified | Page 9

 INTERNATIONAL RESEARCH JOURNAL OF ENGINEERING AND TECHNOLOGY (IRJET) E-ISSN: 2395-0056

VOLUME: 07 ISSUE: 12 | DEC WWW.IRJET.NET P-ISSN: 2395-0072

2020
chatroom)) and MySpace (16346 files, 380,000
comments each (one for each thread)).
Kongregate, an online gaming site, provides
user messages from chat logs. Due to inherent
frustration when playing online games, as well
as a textual way to reach opponents,
aggression is common in the posts. Slashdot is
a discussion-based social networking site
wherein users broadcast messages to others.
MySpace is a popular social networking
website. Datasets are in the form of XML files
each containing and describing a discussion
thread with multiple posts. Each post was
extracted as as angular data element. Each
data element is considered as one document
and indexed through the inverted file index,
assigning an appropriate weight to each
individual term. Applying Lib SVM using a
linear kernel, followed by tenfold cross-
validation gives a false positive of 28 in 294
instances and false negative of 12 in 10184
instances . The model used in this case is the
weighted TF IDF model. Over sampling of the
training cases was used to improve the
training. SVM with linear kernel using
unigrams gives an accuracy of 79.6% while
with bigrams it gives 8l.3% leading to the
conclusion that bigrams should be used with
the SVM linear kernel. This conclusion was
obtained after testing the above on twitter
corpus data (1762 tweets - 39% labeled as
bullying traces). Taking all this data into
consideration our conclusion is that linear
SVM in combination with bigrams gives the
best possible accuracy.[18]

4.3 DNN Model

Deep Layered Network Architecture

Deep neural networks compose computations

performed by many layers. Denoting the
output of hidden layers by h(l)(x), the
computation for a network with L hidden
layers is:

f(x)=f[a(L+1) (h(L)(a(L) (…(h(2)(a(2)(h(1)(a(1)

(x))))))))].

Each preactivation function a(l)(x) is typically a

linear operation with matrix W(l) and bias b(l),
which can be combined into a parameter θ:

a(l)(x)=W(l)x+b(l), a(l)(xˆ)=θ(l)xˆ,l=1 a(l)(hˆ(l-

1))=θ(l)hˆ(l- 1),l>1

© 2020, | Impact Factor value: | ISO 9001:2008 Certified | Page 10

 INTERNATIONAL RESEARCH JOURNAL OF ENGINEERING AND TECHNOLOGY (IRJET) E-ISSN: 2395-0056

VOLUME: 07 ISSUE: 12 | DEC WWW.IRJET.NET P-ISSN: 2395-0072

2020

The “hat” notation xˆ indicates that 1 has been who can then follow-up with appropriate actions Twitter
appended to the vector x. Hidden-layer activation will not allow to go in the profile of user for this we
functions h(l)(x) often have the same form at each might create our own system which can identify such
level, but this is not a requirement. changes and will determine how the bullying affected
person.
In contrast to graphical models such as Bayesian
networks where hidden variables are random
variables, the hidden units here are intermediate
deterministic computations, which is why they are
not represented as circles. However, the output
variables yk are drawn as circles because they can
be formulated probabilistically.

Figure 2: DNN Model[13]

V. Features

5.1 Detection of Non-Textual Cyberbullying

We are going to develop an application which has

image in tweets or online data and we will fetch
such image from twitter and after OCR
classification will be done by our model SVM or
naïve bayes model.

5.2 Expanding Cyberbullying Role Detection beyond

Victims and Bullies

Roles such as instigators, defenders, and

bystanders will be identified by us based on the
algorithm model generated by us by collecting and
labeling such type of data.

5.3 Determining a Victim’s Emotional State after a

Cyberbullying Incident

a victim may change his/her profile details

following such interactions, post content
containing negative sentiments, or leave the
network abruptly. Such instigating interaction can
be flagged up for subsequent review by a human

© 2020, | Impact Factor value: | ISO 9001:2008 Certified | Page 11

 INTERNATIONAL RESEARCH JOURNAL OF ENGINEERING AND TECHNOLOGY (IRJET) E-ISSN: 2395-0056

VOLUME: 07 ISSUE: 12 | DEC WWW.IRJET.NET P-ISSN: 2395-0072

2020
5.4 Word Representation Learning for
Cyberbullying Detection

Experiments can be performed to generate

word embeddings from different datasets,
ranging from general corpora (e.g., Wikipedia)
to more specialised datasets (e.g., abusive
tweets) to compare their effectiveness for
cyberbullying detection.

5.5 Detecting Cyberbullying in Streaming Data and

Real- time

We will determine the cyberbullying on twitter

dataset oauth token will be generated on
twitter account we will fetch the tweets.

5.6 Evaluating Annotation Judgement

We will annotate the each twitter sentence

and output will be generated shown on text.
[15]

VI. Future Modification

The validity and accuracy of the predictive

models to detect cyberbullying on twitter in
this case primarily based on the correct
psychometric categorization of the text.

In future it is intended to improve the system

developed by use more accurate dataset and
to detect the cyberbullying or not. We also
apply other machine learning algorithm and
check the accuracy of models. Higher
accuracy model will help to detect more
accurate bullying.

Another interesting direction for future work

would be the detection of fine-grained
cyberbullying categories such as threats,
curses and expressions of racism and hate.
When applied in a cascaded model, the system
could find severe cases of cyberbullying with
high precision. This would be particularly
interesting for monitoring purposes.
Additionally, our dataset allows for detection
of participant roles typically involved in
cyberbullying.

© 2020, | Impact Factor value: | ISO 9001:2008 Certified | Page 12

 INTERNATIONAL RESEARCH JOURNAL OF ENGINEERING AND TECHNOLOGY (IRJET) E-ISSN: 2395-0056

VOLUME: 07 ISSUE: 12 | DEC WWW.IRJET.NET P-ISSN: 2395-0072

2020
http://www.pcmag.com/article2/0,2817,2388540,00.as
p

Figure 3: Interface of system

VII. Conclusion

The goal of this project is to the automatic

detection of cyberbullying-related posts on social
media. Given the information overload on the web,
manual monitoring for cyberbullying has become
unfeasible. Automatic detection of signals of
cyberbullying would enhance moderation and allow
to respond quickly when necessary. However,
these posts could just as well indicate that
cyberbullying is going on. The main aim of this
project is that it presents a system to automatically
detect signals of cyberbullying on social media,
including different types of cyberbullying, covering
posts from bullies, victims and bystanders.

VIII. References

[1] D. Poeter. (2011) Study: A Quarter of Parents

Say Their Child Involved in
Cyberbullying. pcmag.com. [Online].Available:

© 2020, | Impact Factor value: | ISO 9001:2008 Certified | Page 13

 INTERNATIONAL RESEARCH JOURNAL OF ENGINEERING AND TECHNOLOGY (IRJET) E-ISSN: 2395-0056

VOLUME: 07 ISSUE: 12 | DEC WWW.IRJET.NET P-ISSN: 2395-0072

2020
[2] J. W. Patchin and S. Hinduja, “Bullies move id=299094.299105
Beyond the Schoolyard; a Preliminary Look at
Cyberbullying,” Youth Violence and Juvenile [13] https://www.sciencedirect.com/topics/
Justice, vol. 4, no. 2, pp. 148–169,2006 computer- science/deep-neural-network

[3] Anti Defamation League. (2011) [14] An Effective Approach for Cyberbullying
Glossary of Cyberbullying Detection and avoidance ieee paper
Terms.adl.org.[Online].Available:http://www.adl.or
g/educati on/curriculum [15] Approaches to Automated Detection of
connections/cyberbullying /glossary.pdf Cyberbullying: A Survey ieee paper

[4] N. E. Willard, Cyberbullying and [16] Cyberbullying Detection System on Twitter ieee
Cyberthreats: Responding to the Challenge of paper
Online Social Aggression, Threats, and Distress.
Research Press, 2007.

[5] D. Maher, “Cyberbullying: an Ethnographic

Case Study of one Australian Upper Primary
School Class,” Youth Studies Australia, vol. 27,
no. 4, pp. 50–57, 2008.

[6] D. Yin, Z. Xue, L. Hong, B. D. Davison, A.

Kontostathis, and
L. Edwards, “Detection of Harassment on Web
2.0,” in Proc. Content Analysis of Web 2.0
Workshop (CAW 2.0), Madrid, Spain, 2009.

[7] K. Dinakar, R. Reichart, and H. Lieberman,

“Modeling the Detection of Textual
Cyberbullying,” in Proc. IEEE International
Fifth International AAAI Conference on Weblogs
and Social Media (SWM’11), Barcelona, Spain,
2011.

[8] I. H. Witten and E. Frank, Data Mining:

Practical Machine Learning Tools and
Techniques, Second Edition. San Francisco, CA:
Morgan Kauffman, 2005.

[9] R. Quinlan, C4.5: Programs for Machine

Learning. San Mateo, CA: Morgan Kauffman,
1993.

[10] W. W. Cohen, “Fast Effective Rule

Induction,” in Proc. Twelfth International
Conference on Machine Learning (ICML’95),
Tahoe City, CA, 1995, pp. 115–123.

[11] D. W. Aha and D. Kibler, “Instance-based

Learning Algorithms,” Machine Learning, vol. 6,
pp. 37–66, 1991.

[12] J. C. Platt, “Fast Training of Support Vector

Machines using Sequential Minimal
Optimization,” Advances in Kernel Methods, pp.
185–208, 1999. [Online]. Available:
http://portal.acm.org/citation.cfm?

© 2020, | Impact Factor value: | ISO 9001:2008 Certified | Page 14

 INTERNATIONAL RESEARCH JOURNAL OF ENGINEERING AND TECHNOLOGY (IRJET) E-ISSN: 2395-0056

VOLUME: 07 ISSUE: 12 | DEC WWW.IRJET.NET P-ISSN: 2395-0072

2020

[17] Methods for Detection of Cyberbullying: A

Survey ieee paper

[18] Using Machine Learning to Detect

Cyberbullying ieee paper

[19] Deep Learning Algorithm for Cyberbullying

Detection ieee paper

[20] Online Social Network Bullying Detection

Using Intelligence Techniques ieee paper

© 2020, | Impact Factor value: | ISO 9001:2008 Certified | Page 15