Particle swarm optimization(PSO) suffers from the premature convergence problem in complex optimization problems. To solve this problem, this paper analyzes PSO algorithm from cognitive psychology and proposes a creative thinking(CT) based PSO algorithm(CTPSO). Based on the four stages model in CT process, a framework of CTPSO is designed, and the evolution model is adapted, which includes a CT model besides the memory model, cognitive model and social model in standard PSO to improve the optimization capability of particles. CTPSO is applied to some wellknown benchmarks, and experimental results show that CTPSO possesses more powerful global search capabilities, better convergence rate and robustness, meanwhile it does not introduce new parameters and computational complexity.