【作者】 黄淑丽；

【导师】 王三民；

【作者基本信息】 南昌大学 ， 计算机技术， 2009， 硕士

【摘要】 网络试题库是基于Web的题库系统,该系统由各学科的试题集、题库管理、智能组卷等几部分组成。建设网络试题库的目的是为学校提供一个网络化的题库管理和智能组卷系统。智能组卷是在指定要求下,由计算机从试题库中按照各种约束条件自动组成试卷。如何设计一个好的组卷算法和试题库模型将影响智能组卷的质量和效率。智能组卷系统的目的是智能生成试卷,并能满足或尽可能接近用户预定的目标。因此如何设计一个算法从试题库中高效快速地组成试卷,是文章的研究目的。文章首先分析了试卷的评价指标、各项指标的作用,建立了智能组卷的数学模型;接着论述了通用试题库组卷系统的设计思想及其实现方法并介绍了遗传算法的实现过程。然后,详细介绍了基于传统编码的遗传算法与基于题号编码的遗传算法,在基于题号整数编码遗传算法的基础上提出了基于无重复知识点的遗传算法。该算法将无重复知识点这一刚性约束条件融于遗传算子中,可以简化适应度函数中的约束条件,减少多目标问题中各约束条件之间的竞争,该算法速度更快,组卷质量更好。为验证文章提出的基于无重复知识点遗传算法的可行性,通过改变遗传参数进行了大量的组卷实验,找到了最优参数组合,并验证了算法的有效性。文章还对基于pareto序的多目标优化算法进行了初步的尝试,实现了基于NSGA-Ⅱ的智能组卷算法,取得较满意的效果,同时也验证了NSGA-Ⅱ算法的适用范围。更多还原

【Abstract】 The network item bank is based on the item pool system in the Web form, which consists of question bank of various disciplines , the item pool management ,intelligent group volume. Its goal is to provide a network to the school the question bank management and the intelligent group volume system. The intelligent group volume is under certain request, from tries in the question bank by the computer to be composed the examination paper automatically according to each kind of constraint condition. The group volume question essence is one under the multiple constraint condition multi-objective optimization question.The intelligent group volume efficiency and the quality are decided completely in try the question bank design as well as the group volume algorithm design. Because the intelligent group volume request production examination paper must the greatest degree satisfy the user the different need, therefore how designs an algorithm from to try in the question bank fast to compose the examination paper highly effective, is the article research goal. The article has first analyzed the examination paper appraisal target, each target function, has established the intelligent group volume mathematical model; Then elaborated has tried the question bank group volume system general the design concept and the realization method and introduced the genetic algorithm realization process. Then, introduced in detail based on the traditional code genetic algorithm with based on the topic number code genetic algorithm, in proposed based on in the topic number integer code genetic algorithm foundation based on not duplicates the knowledge genetic algorithm. This algorithm not will duplicate the knowledge this rigidity constraint condition to melt in the heredity operator, might simplify in the sufficiency function the constraint condition, will reduce in the multi-objective question between each constraint condition competition, this algorithm speed will be quicker, the group volume quality will be better. For confirms the article to propose based on not duplicates the knowledge genetic algorithm the feasibility, the article has carried on the massive group volume experiment through the change heredity parameter, has found the most superior parameter combination, and has confirmed the algorithm validity. The article has made an attempt on multi-objective optimization based on non-dominated sorting,realized intellgent test paper generation based on NSGA-II,acquired satisfying effect. At the same time,it validated applying extension of NSGA-II algorithm.更多还原