By integrating chaos mechanism and niche technique, an improved clonal selection algorithm (ICSA) is proposed
based on the clonal selection principle. The algorithm not only maintains better population diversity than the classical clonal selection algorithm, but also converges to the global optimal solution rapidly. By using stochastic processes martingale theory, the martingale characteristic of the average fitness of the population is analyzed and the almost sure strong convergence of ICSA is deduced. Furthermore, it is proved that the algorithm is globally convergent with probability 1 in a finite number of steps when the state space is finite. The simulation results show that ICSA can inhibit prematurity and has preferable global convergent performance.