Detection and analysis of cyberbullying in social media posts with machine learning algorithms

Due to the increasing internet usage in the world and in Turkey, individuals also actively use social media platforms. Users use social media platforms for personal development, shopping, commerce, education, socializing, making friends, producing online content and sharing these contents with other users. But in addition to these well-intentioned uses, there are also uses to harm the other person, humiliate her/him, damage her/his reputation, or knowingly and intentionally bully her/him. The aim of the research is to subject the comments received from Twitter, one of the social media platforms, to natural language processing processes to determine whether there is cyberbullying in the comments and to obtain the best result by comparing the results obtained from the classification algorithms where these data are used as input. In the research, the dataset consisting of 11114 comments was passed through natural language processing processes and given as input data to classification algorithms. Some of these data were used as the training set and some as the test set. As a result, it was observed that the Extra Trees algorithm was more successful than other classification algorithms, with an accuracy rate of 0.87.