教学及考试中的C语言程序题智能评分算法  

Intelligent scoring algorithm of C language program questions in teaching and examination

作　　者：郑玲[1] ZHENG Leng(School of Electronic Information Engineering,Anhui University,Hefei 230601,Aahui,China)

机构地区：[1]安徽大学电子信息工程学院,安徽合肥230601

出　　处：《贵阳学院学报（自然科学版）》2021年第3期99-104,共6页Journal of Guiyang University：Natural Sciences

摘　　要：为有效提高C语言中间表达式解读程序信息的能力,在结合C语言自身语言特点的基础上,引入流程控制图(Flow Control Graph,FCG)匹配自动评分方法,该方法能够根据C语言的特点,计算学生答案与参考答案之间的相似度,给出评分。引入最近邻(Flow Control-KNN,FC-KNN)算法来对FCG算法进行模板脱敏,在FCG的基础上,运用k临近算法根据提取的特征对程序进行评分。实验结果证明,FCG和FC-KNN算法在独立运行时分别具有91.5%和92.3%的平均准确率,而经过融合后,算法之间实现了优势互补,准确率提升到94.0%,在独立运行的情况下,FC-KNN算法的评分效果较FCG好,准确性更高,对两种算法进行数据融合、优势互补,验证了集成后的分类模型在评分的整个过程中,均能够达到良好的分类效果,具有较高的准确率。there are many ways of C language programming,although there are fixed design templates,but due to the diversity of expression,its automatic scoring technology has been a difficult problem.In order to effectively improve the ability of C language intermediate expression to interpret program information,based on the characteristics of C language,flow control graph(FCG)matching automatic scoring method is introduced.According to the characteristics of C language,this method can calculate the similarity between students′answers and reference answers,and give the score.Secondly,the flow control KNN(fc-knn)algorithm is introduced to desensitize the FCG algorithm.On the basis of FCG,the k-nearest neighbor algorithm is used to score the program according to the extracted features.The experimental results show that FCG and fc-knn have an average accuracy of 91.5%and 92.3%respectively when they run independently.After fusion,the advantages of the two algorithms complement each other,and the accuracy is improved to 94.0%.In the case of independent operation,fc-knn has better scoring effect and higher accuracy than FCG.Data fusion and complementary advantages of the two algorithms verify the integrated score Class model in the whole process of scoring,can achieve good classification effect,with high accuracy.

关 键 词：C语言程序题 流程控制图 最近邻算法 集成分类 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]