Online Social Networks News Spreading: Modeling and Data Analysis

The enormous use of social networks in the dissemination of information has made these platforms famous for rumors spreading, but, unlike traditional channels, news is free to propagate without control, so it is of primary importance to understand what are the factors that induce users to propagate a piece of news and not another, and how to identify true and false news. The topic is of interest in multiple areas: psychology, philosophy, politics, marketing, business, finances, and so on. Social networks are constantly modifying the way users create, share and consume information, and were become powerful instruments for understanding social trends and the society behind them, so the purpose of this dissertation is dual. In the first part, after an introduction about complex networks e spreading processes to give the reader the instruments to understand the sequel, I talk about mechanism of social contagion and introduce three different models based on credibility of posts and users who publish news, showing the necessity of an intangible superstructure on the acquaintance network (naturally present in each social network) that take into account this parameter, in order to model the attitude of each user to propagate or not a piece of news. Considering the difficulty to obtain appropriate dataset to test the models, I also implemented three simulators in order to evaluate the influence of the credibility parameters. In the second part, I analyze a dataset of Twitter data containing full cascades of tweets propagated between 2006 and 2016 and concerning specific topics, under the perspective of cascades, writing a mathematical equation that approximate the retweet dynamic, and under the perspective of rumors, finding a method to identify true and false news through the classification of the time series features and the information about users involved in news spreading.