非负矩阵分解基向量分析及其在矿产预测中的应用  

Base vector analysis of non-negative matrix factorization for mineral resources prediction

作　　者：张洁[1] 刘石华 李鸿镇 初晓凤[1] 任雅丽[1] 余先川[1] 

机构地区：[1]北京师范大学信息科学与技术学院,北京100875 [2]广东省地勘局722地质大队,广东汕头440500

出　　处：《地质学刊》2013年第1期81-86,共6页Journal of Geology

基　　金：国家自然科学基金(41072245;40672195)

摘　　要：针对矿产预测数据具有复杂性以及矿质异常信息具有稀疏性的问题,基于非负矩阵分解的非负性和降维的特点,结合稀疏性,提出一种基于NMF基向量分析的矿产预测数据处理方法,并对基向量和原变量以及基向量之间的关系进行分析。广东新寮岽铜多金属矿区数据实验结果表明,NMF方法在不同特征值和相似稀疏度的条件下的基向量形态基本稳定,在保留找矿信息的同时可有效地实现对矿产预测数据的稀疏化。NMF方法对于矿产预测具有重要的实际意义。To solve the problems of the complexity of mineral prediction data and the sparseness of prospecting information, the authors proposed a method of mineral prediction data processing based on the sparseness and the characteristics of dimension reduction of Non-Negative Matrix Factorization （NMF） with consideration of non-negativity, They also analyzed the relationship between base vectors and the relation between vectors and the original variables. The experimental results on the data about Sinliaodong＇s copper polymetallic mine showed that the base vectors of NMF were stable on the condition of similar sparseness and different value of characteristic bases. NMF could realize the sparseness for the mineral prediction data efficiently and retain prospecting information. The experimental results showed that NMF was of important and practical meanings to mineral resources prediction.

关 键 词：矿产预测 非负矩阵分解 化探数据 R型聚类 基向量 稀疏度 广东 

分 类 号：TP392[自动化与计算机技术—计算机应用技术]