高考考生能力数据分析和评价方法研究

发布时间：2018-10-09 21:43

【摘要】：随着互联网技术和信息技术的高速发展,教育领域的学习数据与日俱增。借助于计算机等工具,可以收集到大量的相关数据。如何将这些海量数据转变为知识和信息,并为学习者评价、教学决策及学习优化等方面提供服务,成为教育数据挖掘研究领域的热点。现有的研究主要基于开放和智能的在线学习系统,通过使用学生建模等技术来揭示学生的学习特征和学习状况等。但是,基于作答数据来对学生知识和能力状态的研究较少且更多从心理学的角度进行分析。在心理学领域,认知诊断理论能够了解个体内部微观心理加工过程,实现个体认知情况的诊断。然而,认知诊断在进行诊断分析前一般以Q矩阵理论指导编制测试,这与实际的教育考试存在较大差异。另外,认知诊断模型复杂度较高,且多数模型仅适于0-1评分数据,这使得现有的诊断模型在实际应用中存在一定的局限。针对上述问题,本文基于现有研究提出一种适用于多值评分数据的高考考生能力数据分析和评价方法:首先,基于考生得分信息,通过项目反应理论(IRT)来分析考生在所测项目上的反应模式与其潜在特质(也称能力)之间的关系,对考生个体的能力水平进行评估。具体利用Rasch模型,以极大似然估计方法进行参数估计,计算每个考生个体的总能力值θ。该能力值反映了考生在特定测试上一致性的内在特征,对考生更深层次的能力挖掘有重要作用。其次,由于项目反应理论采用单维性假设,即只适于分析单一维度的能力属性。因此,采用Q矩阵理论将考生不可直接观察的能力状态与在项目上可观察的作答反应相连接,从而把考生能力细化到各个能力属性层面,为考生更深层次的能力分析和评价提供理论依据。具体过程为:(1)详细分析测试项目所测的不可观察的能力属性及其层级结构;(2)利用Q矩阵理论将能力状态转化为可观察的项目反应模式,并改进Q矩阵理论以适于实际考试中的多值评分情况,得到多值评分下的期望反应模式。最后,将考生能力分析和评价看作一个多元分类问题,结合基于IRT和Q矩阵理论的考生能力值和项目反应模式构造考生的特征向量,采用概率神经网络算法进行分类判别,挖掘考生更深层次的能力掌握情况。模拟实验表明,本文提出的结合IRT和Q矩阵的概率神经网络分类方法能够一定程度提高考生能力掌握模式判别的准确率。另外,基于高考数据的实证研究表明,本文方法能够有效实现高考考生能力的分析和评价,并通过不同地区、学校及类别的考生之间的比较分析,能够发现不同群体考生的能力差异,具有一定的实际应用价值。
[Abstract]:With the rapid development of Internet technology and information technology, the learning data in the field of education are increasing day by day. With the help of computers and other tools, a large number of relevant data can be collected. How to transform these massive data into knowledge and information, and to provide services for learners' evaluation, teaching decision making and learning optimization, has become a hot topic in the field of educational data mining. The existing research is mainly based on open and intelligent online learning system, through the use of student modeling technology to reveal students' learning characteristics and learning conditions. However, there are few researches on students' knowledge and ability state based on answer data and more from the perspective of psychology. In the field of psychology, the theory of cognitive diagnosis can understand the process of internal micropsychological processing and realize the diagnosis of individual cognition. However, before the diagnosis and analysis of cognitive diagnosis, the Q-matrix theory is generally used to compile the test, which is quite different from the actual educational examination. In addition, the complexity of cognitive diagnosis model is high, and most of the models are only suitable for 0-1 score data, which makes the existing diagnosis models have some limitations in practical application. In order to solve the above problems, this paper proposes a method to analyze and evaluate the ability of college entrance examination candidates based on the existing research. Firstly, based on the information of candidates' scores, this paper proposes a new method to analyze and evaluate the ability of college entrance examination candidates. Through item response theory (IRT), this paper analyzes the relationship between the reaction mode of the examinee on the test item and its potential trait (also called ability), and evaluates the ability level of the examinee individual. Using the Rasch model, the maximum likelihood estimation method is used to estimate the parameters, and the total ability value 胃 of each examinee is calculated. This ability value reflects the inherent characteristics of the consistency of the examinee in the specific test, and plays an important role in the deeper ability mining of the examinee. Secondly, because the project response theory adopts one dimensional hypothesis, it is only suitable for analyzing the single dimension capability attribute. Therefore, the Q matrix theory is used to connect the unobservable ability state of the examinee with the observable response on the project, so that the examinee's ability can be refined to each ability attribute level. To provide theoretical basis for deeper competence analysis and evaluation of candidates. The specific processes are as follows: (1) the unobservable ability attributes and their hierarchical structure measured by the test items are analyzed in detail; (2) the ability state is transformed into observable item response mode by using the Q matrix theory. The Q-matrix theory is improved to suit the multi-value score in practical examination and the expected response mode under multi-value scoring is obtained. Finally, considering the ability analysis and evaluation of examinee as a multivariate classification problem, combining the ability value of examinee based on IRT and Q matrix theory and item response mode to construct the characteristic vector of examinee, the probabilistic neural network algorithm is used to classify and discriminate. Excavate examinee deeper ability grasps the situation. The simulation results show that the probabilistic neural network classification method combined with IRT and Q matrix can improve the accuracy of judging the ability of examinee to a certain extent. In addition, the empirical research based on college entrance examination data shows that this method can effectively realize the analysis and evaluation of the ability of college entrance examination candidates, and through the comparative analysis of different regions, schools and categories of candidates, Ability to find different groups of candidates ability difference, has certain practical application value.
【学位授予单位】：西南大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP311.13;G632.0

【相似文献】