Mathematical modeling estimates of the reliability of rumors in mass media

Abstract

One of the most important tasks of the contemporary society includes fighting the spreading false information. The unprecedented transition from the traditional media to the modern methods of receiving news has created many problems with verifying its authenticity. Contemporary journalists have to compete with a huge data stream of ordinary users, which is why the main quality factor is the time to publish a news article. As a result, an increasing number of traditional news sources report unclarified information due to the rush to be first. This paper considers a method for determining the presence of hearing in the mass media for the Russian language. This method aims to study the possibility of searching for rumors among users’ messages in social networks. Achieving this goal requires various methods of text analysis, including semantic and linguistic analysis, as well as the analysis of the distribution of records relative to time segments. During the research, the authors have analyzed different popular tools for obtaining data from social networks. In addition, they have manually compiled and marked a sample for training the neural network. As a tool for solving the problem, we used a neural network based on a multi-layer perceptron. The inputs receive a set of 15 metrics that evaluate all aspects of hearing, and as an output, the probability of hearing. The test was performed using various metrics that showed high results for the constructed neural network model. Cross-validation has shown that the model is able to withstand various checks at a high level.