Automated cyberbullying detection using clustering appearance patterns
Walisa Romsaiyud, Kodchakorn Na Nakornphanom, Pimpaka Prasertsilp, Piyaporn Nurarak, Pirom Konglerd · 2017
Cyberbullying is an activity of sending threatening messages to insult person. To prevent cyber victimization from the activity is challenging. This paper enhanced the Naïve Bayes classifier for extracting the words and examining loaded pattern clustering. The algorithm included two main methods: (1) creating partitions by iteratively relocating from entire datasets into clusters using k-mean clustering and (2) capturing any specific partition with the frequency of words with multinomial model feature vector and drawing the probability of words occurring in a document for predicting the eight classes. The proposed method resulted in increasing accuracy and reliability of an experiment.