Understanding the dynamics of information, which shapes many aspects of our social lives, is a major research drive in our increasingly networked society. Be it under the form of a commercial, a rumor, a virus, or a blog post, information diffusion (or propagation) receives substantial attention from multidisciplinary fields of research. A widely accepted approach for exploring the way that intricate relationships between individuals evolve over time, consists of identifying the most powerful, or influential spreaders. Pinpointing influential nodes in real world networks is a considerable challenge, being relevant in interdisciplinary applications, such as information propagation, controlling rumors and disease outbreaks, designing recommender systems, and understanding the organization of social and ecological networks.

The idea of maximizing influence in networks is a nonlinear problem, which represents an element of difficulty for modeling and predicting dynamical systems represented as complex networks . Identifying an optimal, namely minimal set of spreader nodes, remains unsolved despite the vast use of heuristic strategies. Moreover, research on combining the maximization of influence coverage with the minimization of cost of operation is rather scarce and represents a new trend-setter in the state of the art. To tackle this challenge, we propose to find robust trade-off strategies between the cost of maintaining spreaders, the speed of spreading, and the spreading coverage, in order to improve the indoctrination efficiency over a given complex network. Answering these questions can lead to developing a set of ubiquitous strategies for efficient control of information diffusion with direct impact in marketing, sociology and business applications.

As an application of diffusion phenomena, influence propagation in social networks (online and offline) has been studied in various contexts with significant practical potential applications such as viral marketing, monitoring people’s opinions, social psychology analysis and communities discovery. All these interdisciplinary applications are concerned about the role played by a user in a social network and his effect on other users. Current state of the art is limited in terms of methods to identify influential users in social networks - by analyzing their ranking with respect to the users interactivity to the disseminated content - considering both space and time. Furthermore, we consider current solutions using epidemic models of diffusion for validation as having limited performance in the context of social networks. Alternatively, we propose to develop an improved ranking method inspired by a novel benchmarking methodology that makes possible comparison of different centrality ranking methods applied directly on topologies in a competitive scenario, rather than a single-sided scenario, like in the case of epidemiology.

Finally, this project proposes to corroborate its theoretical research results and apply them into an applicative context, namely that of electoral poll forecasting. To this end, we build upon the premises that we can extrapolate the macroscopic opinion dynamics of a society by inferring microscopic temporal dynamic models during the pre-election period. Our hypothesis is that the timing of publicizing opinion polls (i.e., opinion injection) plays a significant role in how opinion oscillates. Research on forecasting election polls was originally constructed by employing classic statistical models, based on so-called macromodels (e.g., national economic and political fluctuations), or micromodels (e.g., surveys of individual voters). Current state of the art in forecasting employs multilevel regression and post-stratification (MRP). However, MRP method is often cumbersome to apply, needing economic indices and detailed demographics to be accurate. Alternatively, we propose to elaborate on the concept of temporal attenuation (TA), which models the timed oscillation of poll data as opinion momentum. For this, we propose a research methodology based on computer simulation of information diffusion, on large datasets, using novel agent-based models, and integrating them with TA in order to improve the forecasting performance of opinion polls.

That the targeted deliverables of this project are:

• D1: one high impact journal publication (in Q1-Q2),

• D2: one journal publication (in WoS JCR),

• D3, D4: two proceedings publications at international conferences,

• D5: a mobile simulator application serving as a proof-of-concept opinion diffusion and real-time poll estimation platform encompassing all the project’s fundamental research outcomes.