Mutation‐based compact genetic algorithm for spectroscopy variable selection in determining protein concentration in wheat grain

Abstract

Wheat is the third most produced grain in the world after maize and rice. Determining the protein concentration in wheat grain is one of the major challenges for measuring its industrial quality. Samples of wheat can be collected using a spectrophotometer device. The challenge is to associate the energy absorbed by the device with the protein concentration in wheat. The device measures hundreds of variable intensities that can be related to the physicochemical properties. The selection of a subset of uncorrelated variables has been shown to be fundamental for establishing correct correlations and reducing prediction error. A new formulation of a compact genetic algorithm that uses only a mutation operator is proposed. The results produced by the proposed approach are compared with traditional techniques for spectroscopy variable selection as successive projection algorithms, partial least square and classical formulations of genetic algorithms. For near‐infrared spectral analysis of the protein concentration in wheat, the prediction errors decreased from 0.28 to 0.10 on average, a reduction of 63%.

A GPU-Based Implementation of the Firefly Algorithm for Variable Selection in Multivariate Calibration Problems

“…The Firefly algorithm was implemented using , and . We have used for RMSEP, MAPE, PRESS, AIC and BIC comparison the Successive Projections Algorithm for MLR (SPA-MLR) [7] , the standard Genetic Algorithm for MLR (GA-MLR) [21] , the Partial Least Squares (PLS) [23] and the Bayesian Variable Selection (BVS). The BVS code was obtained at the following URL: https://github.com/pcarbo/varbvs [24] .…”

“…In GA-MLR, the RMSEP guide the evaluation of a subset of variables used in the calibration model and allows us to choose models more suitable to prediction [21] . Genetic algorithms (GA) are a global search heuristic inspired on the natural evolution of species and in the natural biological process.…”

“…Husselman and Hawick [19] , [20] are the only works we have found so far that present a GPU-based implementation of a FA. Moreover, estimates from the proposed FA (FA-MLR) are compared with predictions from the following traditional algorithms: Successive Projections Algorithm for MLR (SPA-MLR) [7] , Genetic Algorithm for MLR (GA-MLR) [21] , [22] , Partial Least Squares (PLS) [23] and Bayesian Variable Selection (BVS) [24] . In addition, it is used three others criterions to determine the predictive ability of MLR models.…”

Algoritmo do Morcego para Seleção de Variáveis em Calibração Multivariada

“…Com a fi nalidade de comparar os resultados obtidos pelo algoritmo proposto, utilizou-se as implementações propostas em trabalhos anteriores, sendo APS-MLR proposto por Galvão Filho [10], GA-MLR proposto por de Soares et al [12] e FA-MLR proposto por Paula et al [4].…”

Genetic Algorithm for Variable and Samples Selection in Multivariate Calibration Problems

“…The term multivariate calibration refers to the construction of a mathematical model to estimate a quantity of interest on the basis of measured values of a set of explanatory variables (Soares et al, 2014;De Paula et al, 2014;Soares et al, 2010b). Among the traditional technics for construct this model, we can cite the Multiple Linear Regression (MLR) where the data are modelled using linear predictor functions and unknown model parameters are estimated from the data.…”