中文摘要：

为提高题库自动组卷的质量，以ACM Online Judge系统评测数据为研究对象，将时间方差和平均用时作为时空特征对题目进行自动聚类分析；在聚类基础上，使用各分类所有题目的提交次数和提交解决次数计算每类题目的难度系数，并采用高斯随机过程建立自动组卷模型。与传统经验组卷方法相比较，提出的自动组卷模型以题目难度和区分度为依据，组卷质量可科学评价测试者知识水平。实验结果表明，提出的自动组卷模型简单有效，适用性强。

英文摘要：

To improve the quality of auto-generating test paper, a novel method is investigated on testing data from ACM Online Judge system, in which auto-clustering is done on questions by the features of temporal fluctuations and mean of time consumption firstly, then the difficulty coefficient is calculated through the submitting times and solved times of all types of questions, and e- ventually the model of auto-generating test paper is constructed by Gaussian stochastic process. Compared to the traditional mod- el, the proposed mode/can scientifically evaluate testers＇ knowledge level in the light of the difficulty and discrimination of ques- tions. The experimental results suggest that the model is simple but effective and has strong adaptability.