This letter presents a rule-based solution search (RBSS) algorithm for self-optimization in a cellular network. It improves the solution search process of classical optimization methodologies by including a set of rules that capture the knowledge of how a given problem can be solved. The results obtained in a scenario using measurements from a real network reveal that including the RBSS algorithm achieves reductions of up to 60% and 90% in the convergence time of Particle Swarm and Genetic Algorithms, respectively.