This article presents a radio resource optimization model for Radio-over-Fiber (RoF) access networks. The proposed model arranges cells in a multi-tier fashion with increasing coverage radius. Considering the available structure of antennas the optimizer performs dynamic cell merging and cell splitting according to mobile users' demands for efficiently utilizing radio resources. It is proposed an integer programming model with a bi-criteria objective function that tries to minimize the use of network resources as well as to maximize network revenue. The computational demand for obtaining integer solutions increases proportionally to the number of mobile users and also to the number of tiers of antennas. In addition to the integer solutions, an algorithm based on linear relaxation technique is presented, which implies on significant computational time reduction when solving large instances of the problem. Moreover, results are very close to those given by the integer programming formulation.