【作者】 张建国；

【导师】 陈志国；

【作者基本信息】 河南大学 ， 应用数学， 2009， 硕士

【摘要】 智能教学系统中的自动组卷问题实际上是一个在一定的约束条件下多目标参数优化问题,采用传统的数学方法求解此类问题十分困难,自动组卷的效率和质量完全取决于试题库的设计和自动组卷算法的设计。基于项目反应理论的智能考试系统兴起于上世纪八十年代中期,它是智能教学系统及现代测量理论研究中的一个重要领域,它的特点是能够根据考生能力的不同而由计算机从题库中自动选取难度与其能力相匹配的试题进行测试,考试成绩与题目的选取无关,可以更加准确客观地反映考生的实际能力,并且达到一定测量精度要求的时间更短,但是自动组卷时如何将项目反应理论的信息量等计量指标与题型、知识点等非计量指标有机地结合起来,是目前研究的一个热点问题。鉴于以上情况,本文主要做了以下工作:1.分析当前自动组卷算法的研究现状并对常用的几种自动组卷方法加以比较,在充分研究了项目反应理论的基础上,构建了自动组卷问题的数学模型。2.通过对传统遗传算法的学习研究,发现传统遗传算法用在自动组卷时容易出现早熟和收敛速度慢的问题,因此本文对传统遗传算法进行了多处改进,包括编码方式的确定、适应度函数的确定、遗传算子的设计,并且引入了自适应技术和小生境技术。通过用三种不同类型的测试函数对改进后的遗传算法进行了测试,证实了改进后的遗传算法不仅在收敛速度上有了较大提高,而且算法的稳定性也有了显著提高。3.根据自动组卷问题的特点,设计了适当的编码方案和适应度函数,将上述改进后的遗传算法应用于自动组卷问题。以模拟试题库为例进行组卷,并对算法的性能进行了实验分析。实验结果表明,改进后的遗传算法能够成功应用于自动组卷问题,有效地解决了自动组卷中的约束优化问题,具有较高的组卷成功率和效率。更多还原

【Abstract】 Auto-generating test paper is a multi-target parameter optimization question with certain restraint conditions in Intelligent Tutoring System and is difficult to be solved with traditional mathematics methods.The efficiency and the quality of Auto-generating test paper is completely depended on the design of the question bank and the Auto-generating:test paper algorithm.The Intelligent Examination System based on Item Response Theory can be traced back to the middle of 1980’s.It is an important part of the Intelligent Tutoring System and modern measurement research.Its main trait is to construct an optimal test for each examinee,and this is implement by estimating the examinee’s ability.The final grade will be independent of examination items choosing. Intelligence Examination System can be used to measure the real capabilities of the examinees more exactly and more impersonality and quickly.But it brings about a problem in psychometrics that how to comine information quantity with content and other non-psychometrics characteristics.The main contents include:1.The present situation of Auto-generating test paper algorithm and several common methods of Auto-generating test paper are analysed.The mathematical model of Auto-generating test paper question based on IRT is constructed.2.With the research on simple genetic algorithm,the prematurely and the low convergence speed of simple genetic algorithm are discovered.So the paper improves some aspects of genetic algorithms used to deal with the premature and the low convergence speed of genetic algorithm,such as coding strategy, the definition of fitness function,the design of genetic operators and the combination of the Adaptive Technology and the Niching Technology etc.The improved Auto-generating test paper algorithm is tested with three different types of test functions.The test results show that the improved Auto-generating test paper algorithm can improve the convergence speed and the algorithm is quite robust. 3.The proper coding strategy and fitness function are designed according to the traits of automatic group volume question and applies the improved Auto-generating test paper algorithm to the Auto-generating test paper question with the simulation question bank example.The example indicates that the improved Auto-generating test paper algorithm can be successfully applied in the Auto-generating test paper and could solve constraint optimization problems with good performance and practicability.Furthermore,the success rate and efficiency is also high.更多还原