Spectrum classification method based on PCA-UVE-ELM

Abstract

The invention relates to a spectrum classification method based on PCA-UVE-ELM. The PCA-UVE-ELM is a composite algorithm integrating PCA (Principal Component Analysis), UVE and ELM (Extreme Learning Mode); wherein PCA refers to a principal component analysis algorithm, UVE refers to an uninformative variable rejection algorithm, and ELM refers to an extreme learning machine algorithm. Normalization processing is carried out on Raman spectrum original data of a sample, and the normalized data is reduced to intuitive visual projection on a two-dimensional plane by using a PCA algorithm; then preliminary classification is realized on a two-dimensional plane by using a confidence ellipse; the characteristic Raman shift of spectral data of a label is calculated with a relatively high coincidence rate by using a UVE algorithm; enhancement processing is performed on the measured intensity of the characteristic chemical bond Raman shift according to the chemical characteristics of classified substances so as to perform optimization classification, and an ELM model is trained according to a data set according to a ratio of a training set to a test set of 3:1; finally, optimal parameters are searched by using the ELM algorithm, and optimal classification is realized, so that multi-classification of spectrum identification is realized, and the identification and classification efficiency and accuracy are improved.

Description

technical field [0001] The present invention relates to material analysis and identification technology based on Raman spectrum, especially a spectral classification method based on PCA-UVE-ELM. Component analysis algorithm (PCA, principal component analysis, principal component analysis), UVE refers to the uninformative variable elimination algorithm (UVE, uninformative variable elimination, uninformative variable elimination), ELM refers to the extreme learning machine algorithm (ELM, extreme learning machine), through Normalize the raw data of the Raman spectrum of the sample, use the PCA algorithm to reduce the normalized data to an intuitive visual projection on the two-dimensional plane, and use the confidence ellipse on the two-dimensional plane to achieve preliminary classification; The spectral data of the tag uses the UVE algorithm to calculate its characteristic Raman shift, and according to the chemical characteristics of the classified substance, the measured inte...

AI Technical Summary

Problems solved by technology

But the effect of spectral analysis multi-classification problem is not good

Combining machine learning algorithms with spectral analysis is a current research hotspot. Neural networks and support vector machine (SVM, Support Vector Machine) algorithms have good results in Raman spectral analysis. However, due to the high dimensionality of spectral data, data The calculation is cumbersome and the timeliness is poor during the processing

At the same time, the mathematical algorithm is not combined with the physical mechanism of the measured substance, which has certain defects.

Images