Methods for News Items Popularity Estimation on Early Stages

Abstract

Millions of news are distributed online every day. Tools for predicting the popularity of news stories are useful to ordinary people to discover important information before it becomes generally known. Also, such methods can be used to increase the effectiveness of advertising campaigns or to prevent the spread of fake news. One of the important features for predicting information spread is the structure of the influence graph. However, this feature is usually not available for news, because authors rarely post explicit links to information sources. We propose a method for predicting the most popular news in the information flow, which solves this problem by constructing a latent graph of influence. Computational experiments with two different datasets have confirmed that our model improves the precision and recall of forecasting the popularity of news stories.