Opportunistic Mobile Social Networks (OMSNs) are intermittent complex networks consisting of mobile nodes equipped with communication devices, characterized by unstable connections, high node mobility, and limited resources. Existing OSMN routing algorithms, when selecting the next hop node, either focus on limited information within the network and employ simple fusion techniques or consider excessive redundant information in complex designs. To address these issues, this paper proposes an OSMN routing algorithm RFSC based on social centrality and classifier. The algorithm introduces a social circle partitioning algorithm and designs the routing algorithm SCRouter by considering the social attributes of nodes, node-specific situations, and message properties. Subsequently, the SCRouter algorithm is integrated with classifier to enhance the algorithm’s anti-interference capabilities and reduce the consideration of redundant metrics. Furthermore, the RFSC algorithm introduces an adaptive cache replica allocation strategy to improve data transmission efficiency. Experimental results demonstrate that compared to CARA, Prophet, SprayAndWait, CPN_ASW and INR_PRA algorithms, the RFSC enhances delivery rates by 139.6%, 16.6%, 39.5%, 12.0% and 12.7% respectively. It effectively reduces transmission delays and alleviates network congestion.