Abstract: Aiming at the issue that imbalanced data will exist in the snoring patients with Obstructive Sleep Apnea-Hypopnea Syndrome（OSAHS） prediction and analysis by using the data from medical information system. Therefore, on the basis of the clinical research, a screening model based on Random Over-Sampling Examples（ROSE） and C5.0 decision tree algorithm is established. Firstly, to preprocess the collected anthropometry data, remove outliers and fill in the missing value. Then, to apply the ROSE algorithm to balance the preprocessing data. Next, C5.0 classifier is applied to construct the screening model based on the balanced data. Finally, to evaluate the screening effect through 10-fold cross validation. The results demonstrate that this model can effectively improve the screening efficiency of snoring patients with OSAHS.

Key words: imbalanced data, screening model, Random Over Sampling Examples（ROSE）, C5.0 decision tree

摘要： 使用医疗信息系统的数据进行睡眠呼吸暂停低通气综合征（OSAHS）预测和分析过程中，存在不平衡数据问题。为此，在现有临床研究的基础上，提出了一种基于ROSE（Random Over Sampling Examples）和C5.0算法的初筛模型。利用收集到的人体测量学指标数据，通过数据预处理，删除异常值并填补缺失值。然后采用ROSE算法对数据进行平衡，利用C5.0分类器对平衡后的数据构建筛查模型，通过十则交叉验证的方法检验模型的筛查效果。实验结果表明，使用该模型进行打鼾患者的OSAHS筛查，可以有效地提高筛查效率。

关键词: 不均衡数据, 初筛模型, 随机过采样（ROSE）, C5.0决策树