HYBRID MODEL MACHINE LEARNING FOR DETECTING HOAXES

Abstract

Unlimited availability of content provided by users on social media and websites facilitates aggregation around a broad range of people's interests, worldviews, and common narratives. However, over time, the internet, which is a source of information, has become a source of hoaxes. Since the public is commonly flooded with information, they occasionally find it difficult to distinguish misinformation disseminated on net platforms from true information. They may also rely massively on information providers or platform social media to collect information, but these providers usually do not verify their sources. The purpose of this research is to propose the use of machine learning techniques to establish hybrid models for detecting hoaxes. The research methodology used here is a feature extraction experiment, in which a series of features will be analyzed and grouped in an experiment to detect hoax news and hoax, especially in the political sphere by considering five modalities. The outcome of this research indicates that the relation between publisher Prejudice and the attitude of hyper-biased news sources makes them more possible than other sources to spread illusive articles, besides that the correlation between political Prejudice and news credibility is also very strong. This shows that the experiment using a hybrid model to detect hoaxes works. well. To achieve even better results in future research, it is highly recommended to analyze user-based features in terms of attitudes, topics, or credibility.