Toward capturing heterogeneity for inferring diffusion networks: A mixed diffusion pattern model

Highlights：

• Define the problem of inferring diffusion network by thinking of diverse latent diffusion patterns to capture heterogeneity of diffusion processes.

• Analyze massive real-world diffusion data to present an improved pattern-based pairwise transmission probability formulation for diffusion cascade modeling.

• Propose a mixed diffusion pattern cascade model that infers diffusion network by automatically distinguishing latent diffusion patterns.

• Several basic diffusion patterns with distinct spatial, temporal and semantic characteristics are found.