Social Media Forensics: An Adaptive Cyberbullying-Related Hate Speech Detection Approach Based on Neural Networks with Uncertainty

Cyberbullying is a worldwide crisis that affects victims and society as a whole. It is a social media network issue. Because social media platforms are complex and use complex vocabulary, it has become very difficult to automatically detect cyberbullying on these platforms. It might be difficult to precisely grasp the intended meaning of a document because of its casual and concise style, which frequently leads to ambiguous or imprecise language. When confronted with unclear or contextually ambiguous content, identifying cyberbullying becomes even more difficult. There are now many methods for detecting cyberbullying, but they still struggle to differentiate between different types of hate speech linked to cyberbullying because it is unclear and ambiguous and because they are not very accurate. Through the integration of Neutrosophic Logic into the Multi-Layer Perceptron (MLP) model, this paper suggests a novel method for fine-grained cyberbullying classification. By reducing the difficulties caused by the vagueness and overlapping borders between different forms of cyberbullying, the suggested model improves cyberbullying types. By addressing the ambiguity, indeterminacy, and uncertainty inherent in classification decisions, Neutrosophic Logic provides a more thorough and adaptable method for managing intricate categorization scenarios. Because there are overlaps and unclear instances with other types of cyberbullying, the model, which uses the one-against-one technique in MLP classification, captures complex interactions between multiple types of cyberbullying. This model’s testing phase highlights the importance Using class probabilities from several one-against-one classifiers, Neutrosophic Logic offers a thorough understanding of classification results.The results of the proposed model demonstrate the performance enhancement of incorporating Neutrosophic Logic for fine-grained cyberbullying classification tasks.