In this paper, we introduce a radio resource optimization model for Radio-over-Fiber (RoF) access networks. We propose a mixed integer programming model to minimize the network cost while providing service to all mobile users. The proposed architecture arranges cells in a multi-layer fashion, with cells in each layer providing different coverage radius. The optimization algorithm performs dynamic cell splitting to improve network capacity in congested areas and cell merging in areas with low demand to save resources. The computational demand of the proposed model increases proportionally to the number of layers of RAUs in the infrastructure. Results indicate that two layers of RAUs give the best trade-off between network cost reduction and computational demand.