In order to deal with the problems of the slow convergence and easily converging to local optima, a classification learning PSO is proposed based on hyperspherical coordinates. The method of determination of poor performance particle is presented, and the swarm is divided into three parts where three learning strategies are introduced to improve the swarm to escape from local optima. Additionally, to decrease outside disturbance, the particle positions and velocities are updated in hyperspherical coordinate system, which improve the probability flying to the optimal solution. The simulation experiments of three typical functions are conducted, and the results show the effectiveness of the proposed algorithm. Consequently, CLPSO-HC can be used as an effective algorithm to solve complex multimodal problems.