Table 1 Baseline models for information diffusion

For a given hashtag h, M₁ randomly samples the same number of tweets or users as in the real data.

M₂ takes the network structure into account while neglecting social reinforcement and homophily. M₂ starts with a random seed user. At each step, with probability p, an infected node is randomly selected and one of its neighbors adopts the meme, or with probability 1 − p, the process restarts from a new seed user (p = 0.85).

The cascade in M₃ is generated similarly to M₂ but at each step the user with the maximum number of infected neighbors adopts the meme.

In M₄, the simple cascading process is simulated in the same way as in M₂ but subject to the constraint that at each step, only neighbors in the same community have a chance to adopt the meme.