Abstract: A two-point tabu search operator is proposed in order to avoid the iterance during optimization. It is adapted to simulate the activities of a single fish such as pursuing the historically optimal fish, pursuing the current optimal fish, converging to the centroid of the other fishes. The non-uniformity mutation operator used in genetic algorithm is used to represent the individual activity of searching food. The whole fish swarm communicate and cooperate by the above four activities, thereby generating tabu fish swarm algorithm. The tabu fish swarm algorithm is applied to the search of the minimum factor of safety of complicated soil slopes. Comparison of the results by tabu fish swarm algorithm with that by the basic fish swarm algorithm shows that the tabu fish swarm algorithm can find the global optimum and is applicable to constrained optimization problems.