Detection and Control of Cyberbullying via Machine Learning
Kshitij Shanker Rai, Kirandeep Singh, Hritik · 2023
Cyberbullying is disturbing harassment that has grave repercussions. It can be found in several forms and most social networks only display it as text. The online behavior of customers has been addressed, but the developing decade does provide real challenges. It has been challenging to increase incidents of provoking and harassing people as well as casualties. Intelligent frameworks are needed for the programmed discovery of these instances. The majority of the models created in this research may be applied simultaneously to a single social network, and a sizable portion of recent studies have been tackling this issue using traditional machine learning techniques. Machine learning-based models claim they can overcome the drawbacks of conventional models and improve the execution of discoveries, and they have shown methods for identifying cases of online abuse. Although there are numerous antiquated incident control models, an effective torture order is still unnecessary. With the use of machine learning and language preparation, it will be possible to successfully detect harassment in the virtual world and prevent its brutal effects. It is suggested that a methodology be used to describe cyberbullying in two ways. Support Vector Machine, Logistic Regression, the Decision Tree, AdaBoost Classifier, and Random Forest are just some of the machine learning models that have been employed in conjunction with metrics. Precision-Recall Accuracy, F1 score. Logistic regression yielded the greatest value, with accuracy 83.4, precision 84.0, recall 83.2, and F1 score 83.5, and Gaussian nave bayes yielded the lowest value, with an accuracy 45.8, precision 52.2, recall 45.8, and F1 score 45.7.