| 基于改进小生境遗传算法的自适应组卷系统 |
| |
| 引用本文： | 刁建忠,曾爱林,杨斌.基于改进小生境遗传算法的自适应组卷系统[J].顺德职业技术学院学报,2013(4):13-16. |
| |
| 作者姓名： | 刁建忠 曾爱林 杨斌 |
| |
| 作者单位： | [1]顺德职业技术学院电子与信息工程系,广东佛山528333 [2]顺德职业技术学院信息中心,广东佛山528333 |
| |
| 基金项目： | 顺德职业技术学院校级精品课程项目（2010-SZJPKC10）. |
| |
| 摘 要： | 针对传统基于简单遗传算法的组卷系统收敛速度慢，组卷质量较差等缺陷，设计一种基于改进小生境遗传算法的自适应组卷系统。首先依照组卷约束权重比，生成适应度值较高的初始种群．其次对选择策略进行优化。引入小生境预选机制以维持种群多样性，防止过早收敛。再次根据组卷特点改进交叉算子和变异算子，根据种群进化情况，自动调整交叉概率和变异概率。最后经过若干迭代后达到终止条件，以自适应方式生成目标试卷。实验结果证明，与简单遗传算法相比，改进的遗传算法在收敛速度和组卷质量上均有显著的提高。
|
| 关 键 词： | 组卷算法 遗传算法 小生境 自适应 |
Self-adaptive Paper Constructing System Based on Improved Niche Genetic Algorithm |
| |
| Authors: | DIAO Jian-zhong ZENG Ai-linl YANG Bin |
| |
| Institution: | 1. Department of Electronic and Information Engineering, Shunde Polytechnic, Foshan Guangdong 528333, China; 2. Information Center, Shunde Polytechnic, Foshan Guangdong 528333, China) |
| |
| Abstract: | For traditional paper constructing algorithm based on Simple Genetic Algorithm has defects such as low convergence rate, unsatisfactory quality and so on, we designed an efficient paper constructing system based on Improved Genetic Algorithm. Firstly, it generated the relatively higher fitness initial population according to the ratio of paper restraints. Secondly, it optimized the select strategy, imported the niche pre-select mechanism to preserve population＇s diversity and prevented the premature convergence. Thirdly, it improved the crossover operator and mutation operator according to the character of paper constructing, and as the evolution of population carrying on, it could adapt the crossover probability and mutation probability automatically. Finally, it achieved the terminal condition after some iterations, and generated target papers with the self-adaptive method. Experimental results showed that the improved Genetic Algorithm performed much better in convergence speed and paper constructing quality compared to Simple Genetic Algorithm. |
| |
| Keywords: | paper constructing algorithm Genetic Algorithm niche self-adaptive |
| 本文献已被 维普 等数据库收录！ |
|
