Detecting Cyberbullying Behavior in Cyber Data using Bagging Classifier and comparing its Capability over Support Vector Machine Algorithm
P. Manikanta, R. Bhavani, Kavinkumar Krishnaswamy Anbazhagan · 2023
Aim: The suggested research will attempt to perform novel cyberbullying detection by classifying between offensive and non-offensive tweets from Twitter and from its dataset using Bagging Classifier and Support Vector Machine. Materials and Methods: A dataset of abusive and non-offensive tweets is used to develop the Bagging Classifier. Bagging Classifier uses and develops a machine learning method to identify tweets as offensive or not. The sample size was calculated to be 40 per set, and the quality was verified and recorded using Gpower of 80%. Results: The accuracy was maximum in classifications of offensive and non-offensive tweets using Bagging Classifier (94.2%) with a minimum mean error and compared to Support Vector Machine (89.4%). The differences between the classifiers are statistically negligible (p=0.44). Conclusion: In the classification of offensive and non-offensive tweets, the study shows that the Bagging Classifier algorithm outperforms the SVM approach.