Detection of Hate Speech Using Ensemble Models

This study examined cyberbullying on social media sites such as Twitter, with the aim of analyzing these messages before they impact the victims. To achieve this, a global database of 20,010 tweets was used with six different learning machines: LR, RFC, ADB, GNB, SVM, and DTC. The results showed that RFC had the highest F1 (0.90), Precision (0.96), Accuracy (0.91) and highest Re (0.89). These findings raise awareness of the benefits of increasingly popular social media and information sharing. However, it also causes some problems in the spread of hate speech. Researchers are working on using engineering techniques and machine learning algorithms to detect HS in many documents. However, there is currently no research that measures and compares the results of the public information process. This study provides an effective way to detect hate speech on Twitter by using a combination of learning and machine learning techniques. Create a curated database of tweets tagged as hateful, offensive, or neutral to provide a more accurate picture of online discourse. The accuracy of this model is then compared to specific studies to provide a basis for comparison. The results showed that the fused learning method was more effective than the traditional learning method, and RFC showed the highest accuracy in discrimination detection. The research also examines the combination of word embeddings such as Word2Vec and GloVe to improve the performance of the model. Embeddings help improve the model's ability to discern discrimination in online communication. As social media sites continue to serve as vehicles for hate speech and popular information, it is important to develop tools to detect the nature of such speech. With this study, an important step has been taken to achieve this goal. These findings could play an important role in further research to combat hate speech and promote effective online communication.