ANALYSIS OF METHODS FOR DETECTING MISINFORMATION IN SOCIAL NETWORKS USING MACHINE LEARNING

Abstract

Social networks have long become an integral part of the life of modern society. For example, in Ukraine, more than 60% of the population regularly use their functionality. For some people, pages in one or another social network have acquired commercial significance and have become a tool for generating income. There are also rare cases of buying and selling accounts or violating copyright with their help. However, the spread of inaccurate information aimed at misleading and causing serious harm is gaining momentum in social networks. Such a process is defined as “disinformation”.

In addition to disinformation, the term “false information” is also distinguished. These terms are not synonymous, so they should be distinguished for the validity of the study. Misrepresentation is information that contains inaccurate information resulting from errors, but the term does not include the intent to mislead. In turn, the term “disinformation”, on the contrary, is created for the purpose of deliberately spreading false information with the aim of misleading others.

In recent years, the topic of disinformation, as well as its consequences, has attracted a lot of attention. Although disinformation is not a new phenomenon, technological advances have created the perfect environment for its rapid spread. Social networks such as Facebook, Twitter and YouTube create fertile ground for the creation and dissemination of misinformation and false information. This makes it important to research how social media works, how fake news is created and spread through social media, and what role users play.

The study examines social media as a platform for spreading misinformation. Consideration of the problem of user interaction with news in social networks complements the problem of fake news by considering the problem of user interaction with news and collaboration in the information age.

For the reliability of the research, the concepts of misinformation and false information were considered. A comprehensive review of existing approaches to detecting fake news from the point of view of machine learning is given.

Machine learning based classification algorithms play a very important role in detecting fake news or rumors in social media, which is a very complex and difficult process due to various political, socio-economic and many other related factors.

This review covers various machine learning approaches such as Natural Language Processing (NLP), linear regression, k-Nearest Neighbors (KNN), Support Vector Method (SVM), Long Short-Term Memory (LSTM), artificial neural networks and many others.