考试数据分析及孤立点检测的谱聚类方法

Spectral Clustering Method for Exam Data Analysis and Outlier Detection

孤立点(outlier)又称离群点,为位于远离与之相应的随机变量平均值的点。针对考试数据分析和孤立点检测问题,给出了答卷数据的谱聚类算法。利用答卷数据的谱聚类算法对所有考生的答卷数据进行聚类后,根据答卷数据的谱聚类算法中引入的距离矩阵,以每一位被试的答卷数据为参照,构成一个分类图,从而构建一个分类图簇。谱聚类算法在分析考试数据和检测孤立点方面具有明显的优势。实验表明,利用答卷数据的谱聚类算法对考试答卷结果进行聚类和分析,可以从中挖掘出更多有价值的信息。

An outlier is defined as a point that lies very far from the mean of the corresponding random variable. For the problem of exam data analysis and outlier detection, give the algorithm of spectral clustering on exam data analysis. After clustering of all the answers of people who take the exam, by using spectral clustering algorithm of examination scripts, according to the introduction of the distance ma- trix in spectral clustering algorithm, each people who takes the exam can be treated as a reference, to constitute a classification chart, thus build a cluster of classification chart. The algorithm of spectral ,clustering has obvious advantages in the analysis of exam data and outlier detection. Experiments show that clustering and analysing the results of exam answers by using of spectral clustering ,can dig out the more valuable information.