基于主特征映射的多权重数据统计融合可靠性评估模型  

Reliability Evaluation Model of Multi-weight Data Statistical Fusion Based on Principal Feature Mapping

作　　者：董建华 唐家银[1] Dong Jianhua;Tang Jiayin(School of Mathematics,Southwest Jiaotong University,Chengdu 611756,China)

机构地区：[1]西南交通大学数学学院,成都611756

出　　处：《统计与决策》2022年第7期26-30,共5页Statistics & Decision

基　　金：教育部人文社会科学研究规划基金项目(20XJAZH009);西南交通大学新时代“大思政”育人工作项目(DSZ2019-ZLTS-19);西南交通大学本科教育教学研究与改革项目(20201033)。

摘　　要：为解决仅利用单一数据源进行可靠性评估时估计精度不高的问题,文章建立多权重数据统计融合可靠性评估模型。针对大样本情况,采用核密度估计及回归分析法确定密度函数,用权重法对密度函数加权得到合成密度函数,由分布函数的逆函数模拟生成随机样本;在小样本情况下,选用Bootstrap方法估计分布特征参数,通过参数加权进行数据融合;对多维数据借助多维核密度估计进行融合,并在融合算法基础上建立可靠性分析模型。最后,通过算例验证了该类模型的可行性和有效性。In order to solve the problem of low estimation accuracy when is only uses a single data source for reliability evaluation, this paper establishes a reliability evaluation model of multi-weight data statistical fusion. In the case of large samples,kernel density estimation and regression analysis are used to determine the density function. The density function is weighted by the weight method to obtain the synthetic density function, and the inverse function of the distribution function is used to simulate and generate random samples. In the case of small samples, Bootstrap method is used to estimate the distribution characteristic parameters, and data fusion is carried out by parameter weighting. The multi-dimensional data are fused by means of multi-dimensional kernel density estimation, and the reliability analysis model is established based on fusion algorithm. Finally, an example is given to verify the feasibility and validity of the model.

关 键 词：核密度估计 权重法 BOOTSTRAP方法 数据融合 

分 类 号：TB114.3[理学—概率论与数理统计] O213.2[理学—数学]