DETECTION OF CYBERBULLYING WITH CATBOOST CLASSIFICATION AND N-GRAM LANGUAGE MODEL

ADEBAYO, NEHEMIAH PEACE ADEOLUWA (2022) DETECTION OF CYBERBULLYING WITH CATBOOST CLASSIFICATION AND N-GRAM LANGUAGE MODEL. Masters thesis, Landmark University, Omu Aran, Kwara State.

Text Nehemiah Research new new version LMU (updated).docx Download (563kB)

Abstract

The exponential rise in the use of social media and the number of users have led to a proportional increase in the frequency of bullying as social media provides a thriving environment for its perpetrators. Instant messaging platforms such as Facebook, Twitter, and WhatsApp are common areas where the effects of online bullying are well pronounced. The overall negative effects of cyberbullying are not far-fetched as it leads to depression and suicidal tendency among teenage victims. As a result, there is a need to develop measures for detecting cyberbullying, hence, the purpose of this study. Although, in recent years, studies have been conducted to automatically detect instances of cyberbullying on social media with the use of conventional machine learning algorithms. In contrast to other studies, this study is aimed at using the CatBoost classification algorithm in conjunction with the n-gram language model to automatically detect cyberbullying while taking context into account. The language model works by capturing the context, sentiment, and frequency of words from the Twitter dataset. After the training and testing of the CatBoost classification algorithm and the n-gram language model, and other classification algorithm namely Naive Bayes, Support Vector Machine, Random Forest, and Decision Tree, results shows that the CatBoost outperforms other classification algorithm with an accuracy of 97%. In terms of precision and recall, the CatBoost model gave a performance of 90% and 86% respectively when tested with random real-life textual data on the web application framework. The Web application framework developed will aid in detecting instances of cyberbullying in real-life scenarios, thereby reducing the impact of cyberbullying.

Actions (login required)

View Item