【摘要】 高质量的决策越来越依赖于高质量的数据挖掘及其分析,高质量的数据挖掘离不开高质量的数据.在大型仪器利用情况调查中,由于主客观因素,总是致使有些数据出现异常,影响数据的质量.这就需要通过适用的方法对异常数据进行检测处理.不同类型数据往往需要不同的异常值检测方法.分析了大型仪器利用情况调查数据的总体特点、一般方法,并以国家科技部平台中心主持的"我国大型仪器资源现状调查"(2009)中大型仪器使用机时和共享机时数据为主线,比较研究了回归方法、基于深度的方法和箱线图方法等对不同类型数据异常值检测的适用性.选取不同角度,检验并采用不同的适用方法,找出相关的可疑异常值,有助于下一步有效开展大型仪器利用情况异常数据的分析处理,提高数据质量,为大型仪器利用情况综合评价奠定基础,也为科技资源调查数据预处理中异常值检测方法提供有益借鉴.更多还原

【Abstract】 High-quality decisions are increasingly dependent on the quality of data mining and analysis,data mining can not do without high quality data.Instruments used in large-scale survey,due to the subjective and objective factors,resulting in some data always abnormal, affecting the quality of the data.This will require application of different methods for detection of abnormal data processing.Different types of data often requires a different outlier detection method.This paper analyzes the overall characteristics and the general method of the survey data related to the utilization of the large-scale instrument,and the State Science and Technology Center hosted platform," Survey of China’s large-scale equipment resources" (2009) in the data of run time and sharing time of large-scale instrument the main line to the comparative study the statistics of the regression method,the method based on the depth chart and box methods for outlier detection of different types of data applicability.Select a different point of view,test and apply different methods to find related suspicious outliers, the next step will help to effectively carry out large-scale use of unusual instrumentation data analysis and processing,improve data quality,the use of large instruments lay the foundation for the comprehensive evaluation,but also for scientific and technological resources in the pre-survey data outlier detection method provides a useful reference.更多还原