Selection of Near Infrared Spectral Wavelength Variables Based on Improved Immune Genetic Algorithm
  
View Full Text    Download reader
DOI:
KeyWord:near infrared spectrum  wavelength selection  improved immune genetic algorithm  analysis model  prediction accuracy
  
AuthorInstitution
TAO Huan-ming,GAO Mei-feng Key Laboratory of Advanced Process Control for Light Industry (Ministry of Education),School of Internet of Things Engineering,Jiangnan University,Wuxi ,China
Hits: 1131
Download times: 560
Abstract:
      Based on immune genetic algorithm(IGA),an improved immune genetic algorithm(iIGA) was proposed to select the wavelength variables of near infrared spectra.The idea of fixed antibody similarity threshold in the original algorithm was abandoned in the iIGA,which was replaced with adaptive antibody similarity threshold.Meanwhile,the elitist retention strategy and greedy algorithm idea were introduced,making the algorithm carry out local optimization in the right direction.The algorithm was tested on the corn starch and protein content data sets to establish a partial least squares(PLS) analysis model,and compared with IGA,genetic algorithm (GA) and full spectrum method.Results showed that the root mean square error of prediction set (RMSEP) of iIGA was reduced from 0.312 0 to 0.298 0,compared with those of the original IGA algorithm,the prediction accuracy of prediction set was improved by 4.5%.In the prediction of corn protein content,the RMSEP decreased from 0.124 4 to 0.110 3,the prediction accuracy of prediction set increased by 11.3%.A significant test was carried out for the RMSEP values of starch and protein models,respectively,in which F values were 165.22 and 182.05,P values were 9.5 × 10-23 and 4.5 × 10-24,respectively,and P values were less than 0.05.Therefore,iIGA could significantly improve the prediction accuracy of the model.
Close