Towards an Algorithmic Selection of Spreaders in Twitter

No Thumbnail Available
Journal Title
Journal ISSN
Volume Title
Pontificia Universidad Javeriana
This research empirically studies mechanisms and criteria for selecting spreaders in Twitter. Spreaders are users capable of disseminating information to large portions of the network, whether they are considered in uentials or not. This work is an initial approximation in which the Twitter social network is represented as a network and the selection mechanisms exploit its structural properties to nd suitable spreaders. Because the selection depends on di erent mechanisms (e.g. algorithmic, manual, and random selection), a comparison of the distinct features they generate, such as cost and coverage, is investigated. The cost of a spreader is assumed proportional to the coverage potential, that is, the larger the coverage potential, the greater the cost. The cost associated with the selection mechanism is the sum of the coverage potentials of all the selected initial spreaders. This work models, simulates and analyzes how di erent structural properties that characterize the nodes of a network shape information spreading in terms of the coverage and the cost of the initial spreaders. Extensive experimentation is carried out using data from the Twitter social network. These experiments illustrate how di erent selection mechanisms help shaping the dynamics of the spreading process, as well as the cost of the spreaders. Certain network metrics provide good insight for cost-e ective spreader selection, meaning that some metrics (node properties) lead to the identi cation of users with good capabilities to spread information. In general, this work identi es conditions under which an algorithmic selection mechanism o ers the best performance in terms of coverage and cost, and network metrics characterize the optimal initial spreaders in the network. The ndings can o er an alternative approach to select spreaders in commercial and advertising campaigns.