k-means clustering method for coinfective spreading risk calculation

We devise a systematic method for risk estimation of coinfective spreading dynamics, based on k-means clustering. By classifying the outbreaks into high and low risk incidents, we define the mean outbreak size and the outbreak probability.
We simulated the Susceptible-Infectious-Recovered epidemic model on multiple empirical temporal networks (exemplary results in Fig.1-Left and Fig.1-Middle). We classify the realizations, using k-means clustering (Fig.1-Right). By using different shuffling methods we studied each correlation’s effect on hindering/enhancing the spreading phenomena. We observed that causal temporal correlations reduce the size of an outbreak, on the other hand, periodical correlations which can either decrease or increase the probability of an outbreak, have no significant effect on the size of a possible outbreak.