The continuous expansion of new media platforms makes marketing communication present multi-node, multi-path and strong dynamic characteristics. Traditional delivery methods are difficult to take into account coverage, conversion efficiency and resource utilization. This paper proposes a communication path optimization framework for complex networks to maximize marketing effectiveness, which constructs a marketing communication network by using user behavior logs, content features, interaction relationships and conversion feedback. This study identifies nodes with high response, high diffusion and high conversion potential through behavior data preprocessing, user behavior sequence modeling and attention weighting of key propagation features. The divide and conquer algorithm is further introduced to divide the large-scale network into multiple sub-networks, and the global optimal path is generated on the basis of local path search and cross-community bridge fusion. The experimental results show that the coverage rate of the proposed method is 95.1%, the click conversion rate is 8.9%, NDCG@10 is 85.7%, the resource utilization rate is 93.8%, and the average path search time is 13.6 s, which is better than the comparison methods. Research shows that this method can improve the accuracy, stability and computational efficiency of new media marketing communication path selection, and provide reference for intelligent delivery, user hierarchical reach and cross-community communication optimization.
