Machine Learning Methods in Cyberbully Detection: A Concise Review

Cyberbullying is an online crime committed against a victim of harassment and hostility. Despite the development of several techniques for detecting cyberbullying, many studies have mostly relied on textual and specialized user features that were created by the system's designer. This review paper aims to explore cyberbullying detection developed for social platforms using machine learning classifiers, such as Logistic Regression (LR), Random Forest (RF), and Support Vector Machine (SVM). A comparison was also conducted on the performance of the reviewed algorithms in terms of accuracy and precision to determine the classifiers’ recognition rates in identifying cyberbullying on multiple social platforms. Subsequently, the results highlighted the hybrid model as the best method to identify the incidences, whereby the integration of SVM and Convolutional Neural Network (CNN) model. This integration between SVM and CNN has achieved the best accuracy of 89.9% for the detection of cyberbullying texts on social media. Nonetheless, many SVM models were developed to identify cyberbullying activities. Therefore, further research in this area is required to create more precise and efficient methods to identify and prevent online abuse and harassment and improve internet inclusivity among various types of users.