Influence maximization in twitter mention networks: a comparative analysis of PageRank and out-degree heuristics under the Independent Cascade model

Identifying structurally prominent users and selecting effective seed nodes are central problems in social network analysis and influence maximization. This study analyzes a directed Twitter mention network constructed from approximately 1.6 million tweets, comprising 552,121 users and 641,497 directed edges. We characterize user positions using PageRank, in-degree, and out-degree centrality and compare PageRank-based and out-degree-based seed selection under the Independent Cascade (IC) model. The network exhibits highly skewed structural centrality patterns. Seed-set performance is evaluated through 1,000 independent Monte Carlo simulations with a uniform propagation probability of \(p=0.1\) . We report mean spread, standard deviation, 95% confidence intervals, minimum and maximum spread, and zero-additional-activation frequency. Under the adopted graph representation and IC assumptions, the out-degree-based seed set achieves a substantially larger simulated spread than the PageRank-based seed set. This difference is consistent with the diffusion mechanism because activation attempts follow outgoing mention edges, giving high-out-degree nodes more immediate opportunities to propagate. These findings are conditional on the constructed mention graph and the uniform-probability IC model. A mention edge does not necessarily represent the direction of real-world social influence, and the simulations do not incorporate temporal dynamics, content-specific effects, or empirically estimated propagation probabilities. Therefore, the results should not be interpreted as evidence that either heuristic is universally superior for real-world influence maximization. Overall, the study provides a reproducible structural and simulation-based comparison of two seed-selection heuristics while emphasizing the importance of edge interpretation, diffusion assumptions, uncertainty quantification, and empirical validation.