Clustering of Hoaxes on Microblogs Based on User-Based Features as an Intervention for Hoax Dissemination

The proliferation of hoax news, particularly on social media platforms, poses significant risks to public perception and decision-making. This study aims to develop a comprehensive system to detect and label fake news by leveraging data from Turnbackhoax and Kemkominfo. The research begins with web scraping and text preprocessing, followed by assessing document similarity using Term Frequency-Inverse Document Frequency (TF-IDF) and Cosine Similarity. If the similarity between user-inputted news and known fake news exceeds 60%, the news is flagged as potentially fake. If not, the process proceeds to online news scoring and social media scoring, which evaluate the credibility based on time, message, and website credibility, as well as user-based features such as favorites count, followers count, friends count, statuses count, verification status, and years active. Data normalization is conducted using the Min-Max method, and clustering is performed using DBSCAN and K-means algorithms. The clustering results, validated by silhouette scores of 0.7803 for online news scoring and 0.7857 for social media scoring, indicate the effectiveness of the approach. This system, specifically targeting the Department of Information Technology community at the State Polytechnic of Malang, demonstrates a robust methodology for identifying and mitigating the impact of fake news on social media