CYBER-BULLYING DETECTION USING NAIVE BAYES AND N-GRAM

Twitter is the mainly accepted community or social media in these modern days. The users as small children, youngsters to adults, have the responsibility in boosting the attractiveness of twitter. Though, this community or social media could not be alienated commencing the jeopardy of cyber-bullying which is completed habitually by the users, particularly in the tweets and re-tweets. The menace of cyber-bullying is undoubtedly bothered numerous people because of the impact, agony, disappointment, and harassment it has. Therefore, sentiment analysis in twitter can be prepared in regulate and to discover out the bullying scenario in each tweet. Bulling investigation or analysis is a division of Machine Learning and Text mining science that is used to extract, recognize, and cultivate the data. This research used the Naïve Bayes Classification and Gram Model (Uni, Bi, Tri, and N-Gram) model to scrutinize the bullying scenario or sentiment from each and every tweet collectively. However, approximately 1065 records or tweets are analyzed. Thereinafter, pre-processing techniques are enforced and integrated like case folding, stemming, lemmatization, and Bog of Words for feature extraction and well-form structure of tweets. Subsequently, using the Supervised Machine Learning Model i.e. Naïve Bayes and N-Gram the analysis or investigation is performed. Consequently, Naïve Bayes with Uni-Gram achieved 66.77% of accuracy, Naïve Bayes with Bi-Gram achieved 67.29% of accuracy, Naïve Bayes with Tri-Gram achieved 57.86% of accuracy and Naïve Bayes with Ni-Gram achieved 65.09% of accuracy. The average accuracy therefore is approx 64.46%.