Cyberbullying Detection Using Naive Bayes And N-Gram

Abstract

In today’s modern era most social networking sites is twitter. Its user comprises of youngsters to adults and
even children of small age group and they are responsible of elivating the charm of twitter. Though the users of this
microblogging site are sometimes involve in illegal activities which are done consistently by them such as cyberbullying
and cyberstalkingperticularly by tweets and retweets. The hazard of cyberbullying definitely bothers users due to the
haressment and the distress it causes to them. That is why a sentiment analysis can be prepared in the twitter to examine
and in each tweet bullying is regulate. Bullying investigation or bullying analysis is a part of data mining and machine
learning that can be used to extract, acknowledge and cultivate data. To check the cyberbullying in tweets or to perform
the sentiment analysis, this research use naïve bayes classification and gram model (uni,bi,tri,ngarm). In this research
approximately 1065 tweets or records are analysed, after that preprocessing techniques are performed on those tweets
such as stemming, lemmatization,and bag of words. While emanate out feature and after that the analysis or
investigation is performed using model of machine learning such as naïve bayes and n gram. Finally the accuracy that is
achieved by naïve bayes with uni gram is 66.77%, naïve bayes with bigram is 67.29%, naïve bayes with trigram aquired
accuracy of 57.86% and the accuracy that is achieved by naïve bayes with n gram is 65.09%. Hence the average of
accuracy that is achieved is approximately 64.46%