The rapid growth of social media platforms has significantly increased the volume of user- generated content, which has also led to a rise in cyberbullying and hate speech incidents. Detecting such harmful content in real time has become a critical challenge for online safety and digital forensics. This research proposes an adaptive cyberbullying-related hate speech detection approach based on neural networks integrated with uncertainty estimation techniques. The proposed model utilizes deep learning architectures such as Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) to automatically extract contextual and semantic features from social media text data. To improve the reliability of predictions, uncertainty-aware learning mechanisms are incorporated, enabling the model to identify ambiguous or uncertain cases and reduce false classifications. The framework includes data preprocessing, feature extraction, neural network- based classification, and uncertainty evaluation to enhance detection accuracy and robustness. Experimental evaluation on benchmark social media datasets demonstrates that the proposed system significantly improves hate speech detection performance compared to traditional machine learning methods. The results highlight the potential of integrating neural networks with uncertainty modeling to support effective social media forensics and automated cyberbullying monitoring systems.
