摘要
众所周知,地球化学数据携带有众多地质噪音,这些噪音严重影响地球化学数据信息的客观性与可靠性;对于地球化学大数据融合分析而言,确定样品的有效性及变量优选是滤除地质噪音、建立最优样本集合的必要性工作,因而在地球化学大数据处理分析前需首先进行大样本优选,从而更加客观、真实的揭示地球化学大数据信息及相关地质意义。本文以鄂尔多斯盆地北缘1∶20万地球化学土壤测量数据为例,考虑元素之间的地球化学亲和力与组合匹配关系,建立非线性大样本优选模型。具体做法是基于优选后的样品矩阵,将39个元素变量分解成若干独立因子向量,将最优独立因子向量作为元素组合,其向量各分量作为元素变量的权重,依权重大小进行变量优选;优选后的样本集合可以作为该区地球化学数据分析与信息识别的有效地学信息集合,运用这种集合可以有效开展鄂尔多斯盆地外围铀地球化学分析,并为盆地铀资源预测奠定基础。
It is well known that geochemical data carry a large quantity of geological noise, which seriously affect their objectivities and reliabilities. For fusion analysis of geochemical big data, it is the necessity work to filter out geology noise and establish an optimal sample set by determining the effectiveness of the samples and optimizing of the variables. Thus, big sample optimization is needed to carry out before the geochemical data processing analysis, which will more objectively and truly reveal the geochemical data information and geological significance. A set of 1：200,000 scale geochemical soil measurement data in the northern area of Ordos Basin has been taken as a case study, which considers geochemical affinity between the elements and combination match relationship, then establishes a nonlinear large sample optimization model. The specific approach is based on the sample matrix after optimizing, the 39 element variables has been decomposed into several independent factor vectors, and optimal independent factor vector has been used as the element combination. Each component of the vector has been used as the weights of element variables, then the variables have been selected according to the size of the weight. The samples set after optimizing could be used as the collection with effective geo-information of geochemical data analysis and information identification, and this set could be used effectively to carry out the uranium geochemical analysis for the periphery of Ordos Basin, and lays the foundation for uranium resource prediction in the basin.
出处
《岩石学报》
SCIE
EI
CAS
CSCD
北大核心
2018年第2期363-371,共9页
Acta Petrologica Sinica
基金
国家重点基础研究发展规划项目(2015CB453005)资助
关键词
地球化学犬数据
样本优选
非线性元素亲和模型
鄂尔多斯盆地北缘
Geochemical big data
Sample optimization
Nonlinear element affinity model
Northern margin of Ordos basin
