函数型数据异常检测在光谱数据中的应用
Functional Data Outlier Detection for Spectral Data
-
摘要: 光谱分析技术是分析化学中常用的方法之一,也广泛应用于中药分析等领域. 本文用函数型数据分析的方法分析光谱数据,研究针对光谱数据的异常检测方法, 挑选出异常(离群)的样本.基于现有方法, 我们提出``欧加深度检测方法''. 数值模拟结果显示,欧加深度检测方法能较好地挑选出异常情况的样本.本文将欧加深度检测方法和三种已有方法应用于分析73剂中药六混液光谱数据,结果显示, 欧加深度方法能够检测出全部6剂不合格样本, 有较好的应用前景.Abstract: Currently, spectroscopy technology is widely used in traditional Chinese medicine analysis. In this paper, from the functional data analysis perspective, we study outlier detection methods for spectral data, detect outliers, and propose the ``Oja depth detection method''. Simulation studies demonstrate the advantages of the Oja depth detection method. We compare the Oja depth detection method with three existing methods on a Chinese medicine spectral data of 73-dose six-mixture liquid. The results show the proposed Oja depth method is able to detect all six unqualified samples and has the highest accuracy.