This paper presents an ensemble-based multi-classification machine learning approach to detect multiple types of cyberbullying tweets. The study focuses on enhancing a system to identify six types of cyberbullying, using multi-classification algorithms and TF-IDF (bigram) feature extraction. The proposed method achieved high accuracy, outperforming previous experiments on the same dataset. Two ensemble machine learning methods, employing N-gram with TF-IDF feature extraction, demonstrated superior performance in classification. Three popular multi-classification algorithms—Decision Trees (DT), Random Forest (RF), and XGBoost—were combined into two ensemble techniques: stacking and voting. The stacking classifier reached 90.71% accuracy, while the voting classifier achieved 90.44%. The results show that the framework can efficiently detect six different types of cyberbullying with an accuracy rate of 0.9071. The study contributes to the development of ensemble models for detecting multiclass cyberbullying and improves upon existing methods by achieving higher accuracy and efficiency.This paper presents an ensemble-based multi-classification machine learning approach to detect multiple types of cyberbullying tweets. The study focuses on enhancing a system to identify six types of cyberbullying, using multi-classification algorithms and TF-IDF (bigram) feature extraction. The proposed method achieved high accuracy, outperforming previous experiments on the same dataset. Two ensemble machine learning methods, employing N-gram with TF-IDF feature extraction, demonstrated superior performance in classification. Three popular multi-classification algorithms—Decision Trees (DT), Random Forest (RF), and XGBoost—were combined into two ensemble techniques: stacking and voting. The stacking classifier reached 90.71% accuracy, while the voting classifier achieved 90.44%. The results show that the framework can efficiently detect six different types of cyberbullying with an accuracy rate of 0.9071. The study contributes to the development of ensemble models for detecting multiclass cyberbullying and improves upon existing methods by achieving higher accuracy and efficiency.