Utilizing K-Means Clustering for the Detection of Cyberbullying Within Instagram Comments
Ahmad Muhariya, Imam Riadi, Yudi Prayudi, Indrawan Ady Saputro · Ingénierie des systèmes d information · 2023
With the proliferation of social media platforms like Instagram, Twitter, and Facebook, the dissemination of information has undergone a significant transformation.Instagram, distinguished by its emphasis on visual media, has emerged as a platform of choice for photo and video sharing.Despite its popularity, the platform's vast reach renders it vulnerable to malevolent activities, including cyberbullying.While prior research has employed SVM, NBC, C45, and K-Nearest Neighbors for cyberbullying analysis, these studies predominantly focused on Twitter.This paper presents a novel approach, harnessing the power of K-means Clustering to identify instances of cyberbullying on Instagram.In this study, a labelled dataset is gathered and subjected to pre-processing steps, including case folding, tokenization, removal of stopwords, normalization, and stemming.Subsequently, the K-means Clustering algorithm is implemented and evaluated using 10fold cross-validation.The results indicate a threshold value of 1.0, an accuracy rate of 64.25%, a precision of 79.29%, and a recall of 59.88% in categorizing bullying words on Instagram.This research underscores the potential of the K-means algorithm in effectively distinguishing between bullying and non-bullying comments.A notable advancement of this paper is the integration of the two tf-idf weighting methods with the K-means clustering algorithm, thereby enhancing the accuracy in grouping comment data into cyberbullying and non-cyberbullying categories.