Due to a misunderstanding of the idea of freedom speech, the growth of social media generates several problems. One of these problems is cyberbullying, a serious global problem that has an impact on both victims as well as society in general. Cyberbullying can be defined as a form of bullying that occurs across social media platforms. There have been numerous attempts to stop, prevent or lessen cyberbullying in the literature, but because they depend on the interaction of the victims, they are workable. Consequently, it’s essential to identify cyberbullying without the victim’s participation. The purpose of this thesis is to analyze existing machine learning models in the cyberbullying detection field and to demonstrate our approach for cyberbullying detection. Firstly, we will make an introduction into cyberbullying, natural language processing, machine learning and its algorithms. Secondly, we will focus on our approach both theoretical as well as technological and especially in our dataset, machine learning algorithms and results and generally in the whole process of our machine learning model. Finally, we will present our web application CBDA for predicting and recognizing text as cyberbullying or not and its functionality.