Abstract:
Cyberbullying, a pervasive issue in the digital age, poses a serious threat to individuals'
well-being, necessitating advanced technologies for effective detection and mitigation.
This research focuses on the detection of cyberbullying within the context of the Bangla
language, employing a comprehensive approach that integrates deep learning and
traditional machine learning algorithms. The dataset used for this study is specifically
curated for Bangla, ensuring the model's applicability in diverse linguistic scenarios. In
our methodology, we leverage well-established machine learning models, including
Support Vector Machines (SVM), K-Nearest Neighbors (KNN), AdaBoost, Gaussian
Naive Bayes (GNB), Quadratic Discriminant Analysis (QDA), Ridge Classifier (RC),
and Passive Aggressive Classifier (PA). Additionally, we incorporate sophisticated deep
learning models, Bidirectional Long Short-Term Memory (BLSTM) and Recurrent
Neural Network (RNN), to enhance the detection capabilities. The evaluation process
involves employing Bagging Classifiers to assess the performance of each model,
considering metrics such as accuracy, precision, recall, F1 score, and the Area Under the
Receiver Operating Characteristic (ROC) Curve. We visualize the results through
confusion matrices and ROC curves, providing a comprehensive analysis of each model's
effectiveness. The Bidirectional Long Short-Term Memory (BLSTM) emerges as the
frontrunner among the algorithms, exhibiting the highest scores 99.80% accurate across
key metrics.
