Fig. 5: Classification of contagion mechanisms of the #GiletsJaunes Twitter dataset as the function of \(\hat{\phi }\) (x-axis) and \(\hat{\beta }\) (y-axis) parameters.
From: Distinguishing mechanisms of social contagion from local network view
The notation \({d}_{n}^{{\rm {parameter}}}\) represents the nth deciles of the parameter distribution from the #GiletsJaunes dataset from Fig. 4. The classification results of each instance i are shown at the corresponding location of the decile of its inferred \({{\hat{\phi}}_{i}}\) and \({{\hat{\beta}}_{i}}\) parameters sampled from the \(P(\hat{\phi })\) and \(P(\hat{\beta })\) distributions. The background colour of each panel indicates the dominating classified mechanism that characterise the given parameters (purple for Sm, orange for Cx and blue for Sp). The certainty of classification, displayed with black circles, is defined as the proportion of trees in the random forest that have classified an instance into the assigned contagion type, averaged over the set of instances classified in that contagion type. Most of the infection cases are classified as simple if their \(\hat{\beta }\) are in the 8th decile or below and their proportion of infected neighbours is >\({d}_{5}^{\phi }\), and as complex otherwise.
