Abstract
In today’s digital space, there is a vast exchange of information, particularly on social media platforms, where the spread of harmful content related to violence, racism, bullying, and extremism has become a pressing issue. Such information negatively impacts society, contributes to the radicalization of views, and poses significant security threats, making this issue increasingly relevant. Consequently, the automatic detection and filtering of such content has become one of the primary tasks for big data companies, law enforcement agencies, and social media platforms. In this study, we implemented multi-class text classification using the StackingClassifier, Gradient Boosting Decision Trees, Neural Networks, Gradient Boosting Decision Trees + Neural Networks, GBDT + Neural Networks, Ensemble Learning and GridSearch models. This method combines multiple machine learning algorithms, leveraging their strengths to improve classification accuracy.