Model-oriented fake news detection on social media
Nowadays, fake news (FN) have actively penetrated throughout the social media reducing our ability to critical assess and proceed the information. Most of existing approaches to handle with FN require a labeled FN training datasets but in some cases these datasets are unavailable. In this paper, we present a model-oriented approach for FN detection and feature extraction. The unsupervised technique for FN identification without the training data is designed and developed. It includes four main steps, namely data preprocessing, text feature extraction, vectorization, and clustering using k-means algorithm. The results of the last step was evaluated through several parameters: homogeneity, completeness, V-measure, Adjusted Rand index and Silhouette coefficient.