关键词: 遗传算法, 自动组卷, 小生境, 自适应

Abstract: In order to avoid premature convergence and low convergence speed，an improved adaptive genetic algorithm based on niches is presented according to the feature of population fitness distribution.The algorithm adopts a new method that simulated niche is introduced in the selection operator for the population selection and makes the crossover probability and mutation probability adjust adaptively and nonlinearly.The test results show that the improved algorithm can evidently improve success ratio convergence speed and solve the problem of auto-composing test paper more effectively.

Key words: Genetic Algorithm（GA）, auto-composing test paper, niche, adaptation