Volume: 12 Issue: 05 | May 2025 www.irjet.net
Cyberbullying Detection with Machine Learning Techniques
Y Jyothika1, Likitha R2, Soundarya H J3 ,Dr. J Amutharaj4
Abstract – The rise of social media platforms has inadvertently facilitated the spread of cyberbullying, affecting numerous young individuals. As these platforms proliferate, so does the prevalence of online harassment. This study introduces a machine learning-based approach to detect cyberbullying on social media, particularly Twitter. By leveraging Natural Language Processing (NLP) to analyze textual data and Long Short Term Memory (LSTM) networks for image recognition, the proposed model aims to identify bullying content effectively. The system utilizes Twitter's API to retrieve tweets, which are thenprocessedusingclassifierslikeXGBoostandDecision Trees to determine the presence of bullying behavior. However, many social media bulling detection techniques have been implemented, but many of them were textual based. The objective of our project work is to show the implementation of NLP and LSTM which will identify and classify tweets, posts, and other content associated with bullying. Accordingly, machine learning model is proposed to detect and prevent bullying on Twitter. Two classifiers i.e. NLP(Natural Language Processing) are used for identifying the complete sentence in the comments and LSTM(Long Short Term Memory) for image identification. Both NLP and LSTM were able to detect the true positives with more accuracy. Also, Twitter API is used to fetch tweets and tweets are passed to the model to detect whether the tweets are bullying or not along with we use XGBoost and Decision Tree algorithms.
Key Words: Cyberbullying, NLP(Natural Language Processing),LSTM(LongShortTermMemory),TwitterAPI, XGBoost,DecisionTree.
1.INTRODUCTION
Cyberbullying, a modern manifestation of harassment, has become increasingly common with the advent of digital communication platforms. Unlike traditional bullying, cyberbullying allows perpetrators to remain anonymous, makingitmorechallengingtodetectandaddress.
Forinstance,suspectsinseveral recenthate-relatedterror attacks had an extensive social media history of hate related posts, suggesting that social media contributes to theirradicalization.
Around 87 percent of the today’s youth have witnessed some form of cyber bullying. Cyber Bulling can take different structures like Sexual Harassment, Hostile
Environment, Revenge, and Retaliation. Since offender is hiddentothevictim,theproblemstatementgetscomplex. With the multiplication of online life and internet access, the actofcyber bulling too hasincreased, and it’s difficult to detect. Thus, it is necessary to detect cyber bullying in order to protect adolescents. In this research, this vital data is utilized and information in the form of texts to improvetheexistingcyberbullyingdetectionperformance.
To address this, we propose a system which employs natural language processing techniques and the classificationisdoneusingmachinelearningapproachthat incorporatesvariousclassificationtechniques.
1.1
Purpose and Scope
The purpose of this project is to provide a desktop UI applications for classification of hate speech. To identity the maximum number of hate speech related tweets from twitterassoonasitispostedbyusers.Theproblemcanbe framed as a multi-class classification task, where tweets are categorized into three distinct classes: hate speech (HS), nonhate speech (NHS), or offensive content. This application can be used to classify the hate speech (HS), not hate speech (NHS) or offensive on social media network.
1.2 Problem Statement
In today's digital era, cyberbullying poses a significant threat, leading to emotional and psychological distress amongusers,especiallyadolescentsandyoungadults.The challenge lies in the subtlety and anonymity of such acts, makingmanualdetectionarduous.
. To address this problem there is a critical need for an automated system that can effectively detect and mitigate instances of cyberbullying on social media platforms, chat application,andotheronlinecommunicationchannels.The goal of this project is to develop a robust and accurate cyberbullying detection system using deep learning techniques, specifically long short-term memory(LST) and logistic regression. This system should be capable of analysing text and multi media content( such as images and videos) to identify and classify instances of cyberbullying, hate speech, or offensive content in real time.
International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056
Volume: 12 Issue: 05 | May 2025 www.irjet.net p-ISSN:2395-0072
2. Literature review
1. MelNet: A Generative model for audio in the frequency domain
InthispaperaMelNetmodel,agenerativemodel forspectralrepresentationofaudioisintroduced MelNet combines a highly expensive autoregressivemodelwithamultiscalemodelling scheme to generate high resolution spectrograms generated by MelNet exhibit a realistic structure at both local and global levels. Unlike previous approaches that model time-domain signals directly,MelNetisespeciallyeffectiveatcapturing long-range temporal dependencies. Experimental results indicate promising outcomes across various tasks, such as unconditional speech generation, music synthesis, and text-to-speech conversion..
2. MMM: Exploring conditional multi-track music generation with the transformer
In this paper a novel generative model for multitrack sequence generation under the frame work of GANs. We have also implemented such a model with deep CNNs for generating multi-track piano-roles.Wedesignedseveralobjectivemetrics and showed that we can gain insights into the learningprocessesviathisobjectivemetrices.The objective metrices and the subjective user study showsthattheproposedmodelscanstarttolearn something about music. All though musically and aesthetically it may still fall behind the level of human musicians, the proposed model has a few desirableproperties.
3. Counterpoint by convolution
The paper presents the process of placing notes against notes to construct a polyphonic musical piece. This is a challenging task, as each note has strong musical influences on its neighbors and notes beyond. Human composers have developed systems of rules to guide their compositional decisions. However, these rules sometimes contradicteachother,andcanfailtopreventtheir users from going down musical dead ends. Our current focus on statistical models of music represents one of several computational methods that enable composers to experiment with ideas more efficiently, thereby lowering the costs associated with creative exploration. Whereas previous work in statistical music modelling has reliedmainlyonsequence modelssuchasHidden Markov Models and Recurrent Neural Network (RNNs), we instead employ convolution neural
networks due to their invariance properties and emphasisoncapturinglocalstructure
4. Cyberbullying detection in social network: a comparison between machine learning and transfer learning
Itexplorestheeffectivenessoftraditionalmachine learning versus transfer learning models in identifying cyberbullying content on social media platforms. The research evaluates multiple algorithms using benchmark datasets, highlighting that transfer learning approaches –particularlythoseleveragingpre-trainedlanguage models – consistently outperformed traditional methods in accuracy and contextual understanding. The study emphasizes the importance of deep semantic analysis in addressing online abuse and contributes valuable insights towards building more effective and intelligentcontentmoderationsystem.
5. The role of artificial intelligence and cyber security for social media
ItexamineshowAI technologies can be leveraged to enhance cybersecurity in social media platforms.ThestudyexploresthedualroleofAIin both enabling smarter threat detection (such as identifying fake news, bots, and malicious content) and reinforcing security mechanisms through automated responses and real-time monitoring. It emphasizes the growing need for intelligent systems to combat evolving cyber threats, while also addressing ethical and privacy concernsassociatedwithAI-drivensurveillance.
3. Proposed System
The proposed system integrates machine learning and Natural Language Processing (NLP) methodologies to detectoffensivecontentinonlineinteractions.Itbeginsby collecting and preprocessing data from social media platforms, chat applications, and public databases. To categorize information as bullying or non-bullying, it first gathers and preprocesses data from public databases, messaging apps, and social media. In the process of identifying sarcasm, irony, and emotional nuance-all these are frequently employed in cyberbullying-the algorithm also integrates contextual awareness. When a message is detected as potentially dangerous, it is blocked and automatically responds with reports or warnings. This method guarantees real-time, scalable detection on multiplewebplatforms.Naturallanguageprocessing(NLP) and automated cyberbullying detection are combined in thesuggestedmethod.
International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056
Volume: 12 Issue: 05 | May 2025 www.irjet.net p-ISSN:2395-0072
Advantages of Proposed System
EfficiencyandScalability:
Thesystem’sabilitytoprocessmassiveamountsofdatain real-time allows for quick detection and intervention across the multiplicity of online platforms, that makes themeconomicalandscalable.
Contextualandnuancedrecognition:
The system can precisely detect subtle types of cyberbullying, such as sarcasm and irony, by utilizing sophisticated natural language processing(NLP) and machine learning techniques (e.g., sentiment analysis, sarcasmrecognition,BERT,andLSTM).
Regular feedback and retraining improve the system’s accuracy over time, allowing it to adjust to new language trends and bullying strategies. Automated detection guarantees the consistent and objective identification of harmfulinformation.
4. Methodology
Data collection using twitter tweets:
The sentiment/tweets are collected from a set of 20 accounts. The data retrieval is done by using twitter API using OAuthapi used to authenticate the open-source frameworkwiththetwitterapplication.
Sentimental storage based on tweets:
The sentimental storage based on tweets is a process of storingdataaboutthetweetsintotherelationalstoragein terms (TwitterId, TwitterDesc, UserId). Twitter Id is uniqueIdassociatedwiththetweet,TwitterDescisactual tweetandUserIdistheIdassociatedwiththeuser.
Stop words:
These are the set of words which do not have any specific meaning. The data mining forum has defined set of keywords. Stop words are words which are filtered out before or after processing of natural language data (text). There is not one definite list of stop words which all tools useandsuchafilterisnotalwaysused.
Data cleaning:
Data cleaning is used for removing the stop words from each of the tweets and clean them. After the cleaning processiscompleted,thecleandatacanberepresentedas aset(CleanId,cleandata,userId).CleanIdistheuniqueId associated with the Tweet, Clean Data is the clean data after removal of clean data and user Id is the unique Id associatedwiththeuser.
Music Generation Model (MG):
The Music Generation Model takes into account both the userinputandthedeletedsentimenttodynamicallyadjust the generation of music. The model is extended to be sentiment-aware, considering sentiment information during the composition process. It is trained on a dataset of diverse musical compositions, possibly including sentiment-taggedexamples.
Real-time Preview:
This component provides users with an immediate preview of the music generated based on their input and the deleted sentiment. Users can assess the emotional resonance and characteristics of the music in real-time, influencingtheirfutureinteractions.
User Feedback Mechanism:
Thefeedback mechanismallowsusersto provide input on the generated compositions, sharing their thoughts and preferences. It could include a form within the GUI where users submit comments, ratings, or other feedback. User feedback is essential for refining the sentiment to music mapping and improving the overall system based on user preferences.
Feedback Submission System:
This system is responsible for collecting and submitting user feedback received through the GUI. It collects feedback submissions, which may include qualitative commentsandquantitativeratings.
Feedback Processing:
This component processes the submitted feedback, extracting valuable insights and patterns from user responses. It involves analyzing feedback to understand user preferences, satisfaction, and areas for improvement. The processes feedback contributes to iterative updates andimprovementsinvariousaspectsofthesystem.
Music Player:
The music player component plays the final generated compositions for the user to listen to and evaluate. It produces the auditory output based on the dynamically adjustedmusicgeneratedbytheMusicGenerationModel.
International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056
Volume: 12 Issue: 05 | May 2025 www.irjet.net p-ISSN:2395-0072
Chart -1:Methodology
5. CONCLUSION
This study presents a comprehensive approach to detecting cyberbullying by utilizing diverse datasets and deep learning techniques for feature extraction. The implementation of LSTM networks, enhanced by ReLU activation functions, has shown improved performance overtraditionalmethods.
LSTM performance was improved over sigmoid by ReLU activation.Tofurtherenhancedetectioncapabilities,future research proposes to include picture, video, and multilingualdatasets.
The utilization of Long Short-Term Memory (LSTM) networks for cyberbullying detection represents a promising avenue, with current models demonstrating commendable performance in distinguishing harmful onlinebehavior.Astechnologyprogresses,thefutureholds excitingpossibilitiesforthefield,includingtheexploration ofadvancedneuralarchitectures,multimodalanalysis,and real-time detection. Additionally, the development of context-aware models, personalized approaches, and the integration of behavioral analysis will contribute to more nuanced and accurate detection systems. However, ethical considerations, such as user privacy and fairness, must be at the forefront of development efforts. A collaborative, global approach involving researchers, policymakers, and educators is imperative to tackle the multifaceted challenges of cyberbullying effectively. As technology and society continue to evolve, the ongoing commitment to innovation,education,andethicalstandardswillbecrucial in creating robust and responsible solutions for the detectionandpreventionofcyberbullying.
REFERENCES
[1] Elaheh Raisi Bert Huang., “Cyber bullying Identification using Participant-Vocabulary Consistency”Virginiatech,Blacksburg,VA,2016.
[2] Nebrase Elmrabit Feixiang Zhou, Fengyin Li, Huivu Zhou., “Evaluation of Machine learning Algorithms for AnomalyDetection”2018.
[3] Bhavani Thuraisingham., “The Role of Artificial Intelligence and Cyber Security for Social Media” Computer Science Dept. The University of Texas at Dallas Richardson, USA bxt043000@utdallas.edu 2020.
[4] Zaheer Abbass, Zain Ali, Mubashir Ali, Bilal Akbar Ahsan Saleem., :A Framework to Predict Social Crime through Twitter Tweets By Using Machine Learning” DepartmentofcomputerScienceUniversityofLahore, Guiratcampus,Pakistan2020.
[5] Rahul Ramesh Dalvi, Sudhanshu Baliram Chavan, Aparna Halbe., “Detecting A Twitter Cyber bullying Using Machine Learning” Department of Information Technology Sardar Patel Institute of Technology Mumbai,India2020.