论文部分内容阅读
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.
High-quality decision-making increasingly relies on high-quality data mining and its analysis, and high-quality data mining can not do without high-quality data. In large-scale instrument utilization surveys, some data is always caused by subjective and objective factors Anomalies, affecting the quality of the data.This requires the adoption of the method of detection of abnormal data.Different types of data often require different methods of detection of outliers.This paper analyzes the general characteristics of the survey data of large-scale use of the instrument, the general method, and to National Center for Science and Technology, hosted by the platform “China’s large-scale instrument resources survey ” (2009), large-scale equipment and machine-time sharing data as the main line, a comparative study of regression methods, depth-based methods and box line method Applicability of different types of data for outliers detection Select different angles, test and adopt different applicable methods to find out relevant suspicious abnormal values, which is helpful to effectively carry out the analysis and processing of abnormal data of large-scale instrument utilization in the next step and improve the data Quality, laying the foundation for the comprehensive evaluation of the utilization of large-scale equipment, but also for the preprocessing of scientific and technological resources survey data Value detection method provides a useful reference.