Electrical Electronics Engineering: Open Access
ISSN: 3071-0774 | DOI: 10.64030/3071-0774
M Sadigzade and E Nasiboglu
With the development of the social network, the number of cyberbullying started to increase in the world. Cyberbullying detection is receiving increasing attention, especially in Machine Learning communities. The reason for the increase in cyberbullying is that the bullying on the Internet cannot be detected or even if they are detected, they think that legal sanctions will not be applied. These types of cyberbullying crimes leave mental scars in their lives in the future by putting psychological pressure on people. It is very difficult to identify and counter cyberbullying in a timely manner.
Cyberbullying will be a growing problem in Turkey as in the rest of the world. By the findings so far, 20% are already becoming cyberbullies in Turkey. In this regard, there are few studies in the literature on the detection of cyberbullying in Turkish texts. Machine learning is also being used in ongoing research to detect and eliminate cyberbullying. Although there is a lot of cyberbullying detection in English, there is little research in Turkish. Moreover, only a limited number of algorithms and methods have been used in Turkish studies.
Moreover, the aim of this study is to use different machine learning algorithms to detect Turkish cyberbullying messages. In this study, those who made their quartet drawings on a dataset consisting of 3000 Turkish social networks using cyber techniques. Precision, accuracy, cross- validation, recall and F1 scores were used to appraise the performance of the classifiers. In the study, Linear SVC performed best Train Models for CountVectorizer, with cross-validation score of 89.92% and F1 score of 99.96%, and Linear SVC performed best Train Models for TfidfVectorizer, with cross-validation score of 89.79% and F1 score of 99.96%.