Cyberbully and its impact have occurred around the world and now the number of cases is increasing. The detection of cyberbullying is very important because the amount of information on internet is too large and humans are not able to track all this information.

The purpose of this research is to create a classification model with high accuracy, we built different models like LSTM, CNN and BERT to address the identification and classification problem of recognize 6 classes of cyberbullying. Two word embeddings are used as input of the models, one from spaCy and one created by us using only the dataset.

Finally we obtained the best results with an Ensemble of our best models, with an accuracy of 0,8633.