基于双向检验的异常数据剔除与修复方法

Abnormal Data Eliminating and Repairing Method Based on Two-Sided Test

针对靶场复杂的试验环境及测量数据的特点,深入分析了异常数据产生的原因,提出了一种利用多项式拟合对测量结果数据进行双向检验,剔除异常数据的方法,并在检验结果满足修复条件时,对剔除数据进行加权修复。实际数据测试表明:本方法能够有效克服数据突变和段落性阶跃等干扰因素的影响,具有较好的模型稳定性和较高的异常数据剔除率,满足试验任务实际需求,具有较高的工程应用价值。

In view of the complex test environment and characteristics of measurement data, through a deep analysis of the causes of abnormal data, a method for eliminating abnormal data based on two-sided test of measurement result data by use of polynomial fitting was presented. When the test result met the repairing condition, weighted repairing was conducted for the eliminated data. Real data test showed that this method can effectively overcome the influence of data mutation and paragraph step, has better model stability and higher abnormal data eliminating rate, which can meet the actual requirement of experimental task and is of high practical value in engineering.