Abstract :

Cyberbullying On Social Media Is Common And Can Have Serious Mental And Emotional Effects. It Is Very Important To Have Effective Ways To Find Out About It. Using Advanced Machine Learning And Sentiment Analysis Together Is A Scalable Way To Find Harmful Interactions And Make Digital Places Safer. The Cyberbullying Tweets Dataset From Kaggle Is Used In This Project. It Has Examples Of Different Types Of Cyberbullying That Have Been Named. Normalization, Noise Removal Like URLs And HTML Tags, Tokenization, WordNetLemmatizer, VaderSentiment Analysis, And Label Encoding Are All Parts Of Preprocessing. TF-IDF Vectorization Is Used To Retrieve Features, And SMOTE Oversampling Is Used To Fix Class Imbalance. Tools For Visualizing Data, Like Distribution Plots And Word Clouds, Can Help You See Patterns And Trends In Bullying. The Suggested System Uses Many Different Classifiers, Such As Logistic Regression, Random Forest, XGBoost, Decision Tree, Naive Bayes, SVM, Extra Tree, Gradient Boost, And AdaBoost. It Also Lets You Tune The Hyperparameters Using GridSearchCV. Evaluation Uses Memory, Accuracy, Precision, F1-score, And Confusion Matrices To Catch False Positives And False Negatives. Some More Improvements Are SMOTEENN, Which Balances The Data Better, And A Voting Classifier Ensemble, Which Combines MLP, Bagging, And Logistic Regression For More Accurate Classification. Explainable AI Techniques Like LIME And SHAP Make Sure That The Results Can Be Understood By Finding The Most Important Features. Flask-based Deployment Allows Real-time Predictions With Confidence Scores And Topic Modeling To Make The System Easy To Use And Clear. The Results Show That The Proposed Voting Classifier Does Much Better Than All Baseline Models, Scoring 97.4% Across All Evaluation Measures. This Shows That The System Is Reliable, Scalable, And Useful For Finding Cyberbullying.

Published:

08-4-2026

Issue:

Vol. 26 No. 4 (2026)

Page Nos:

1667-1679

Section:

Articles

License:

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

How to Cite