The Fireworks Model describes the behavior and the time-evolution of signed networks generated by two-sided conflicts. We assumed that links form ex-novo at each step and decay in the next one, making it unnecessary to study the actual edge-dynamics. If this assumption holds, our model reduces the study of these networks to a series of sampling problems. We tested the model on two different datasets: the USPTO patent-litigation dataset and an Interstate-War dataset. We focused on reproducing the main features of the networks, with particular regard to the largest connected component and the number of triangles.
