Predicting the metabolizable energy content of corn for ducks: a comparison of support vector regression with other methods

Abstract

IntroductionCorn (Zea mays L.) is the main energy source in diets for intensively reared avian species (broilers and ducks), therefore accurate information on its effective energy content is of importance to nutritionists. A number of studies have been conducted to estimate the metabolizable energy (ME) content of corn based on its physical characteristics and chemical composition (e.g. Leeson et al., 1993;Zhao et al., 2008). The energy content of feedstuffs depends strongly on their chemical composition. Nutritionists are interested in using models that predict the nutritive value of poultry feedstuffs accurately. Recently, artif icial neural network (ANN) models have received attention among poultry nutritionists, e.g. for estimating the ME of poultry offal meal (Ahmadi et al., 2008) and sorghum grain (Sedghi et al., 2011) AbstractSupport vector regression (SVR) is used in this study to develop models to estimate apparent metabolizable energy (AME), AME corrected for nitrogen (AME n ), true metabolizable energy (TME), and TME corrected for nitrogen (TME n ) contents of corn fed to ducks based on its chemical composition. Performance of the SVR models was assessed by comparing their results with those of artificial neural network (ANN) and multiple linear regression (MLR) models. The input variables to estimate metabolizable energy content (MJ kg -1 ) of corn were crude protein, ether extract, crude fibre, and ash (g kg -1 ). Goodness of fit of the models was examined using R 2 , mean square error, and bias. Based on these indices, the predictive performance of the SVR, ANN, and MLR models was acceptable. Comparison of models indicated that performance of SVR (in terms of R 2 ) on the full data set (0.937 for AME, 0.954 for AME n , 0.860 for TME, and 0.937 for TME n ) was better than that of ANN (0.907 for AME, 0.922 for AME n , 0.744 for TME, and 0.920 for TME n ) and MLR (0.887 for AME, 0.903 for AME n , 0.704 for TME, and 0.902 for TME n ). Similar findings were observed with the calibration and testing data sets. These results suggest SVR models are a promising tool for modelling the relationship between chemical composition and metabolizable energy of feedstuffs for poultry. Although from the present results the application of SVR models seems encouraging, the use of such models in other areas of animal nutrition needs to be evaluated.Additional key words: maize; poultry; nutritive value; chemical composition; artificial neural network; multiple linear regression. Abbreviations used: AME (apparent metabolizable energy); AME n (apparent metabolizable energy corrected for nitrogen); ANN (artificial neural network); CF (crude fibre); CP (crude protein); EE (ether extract); ME (metabolizable energy); MLR (multiple linear regression); SVM (support vector machine); SVR (support vector regression); TME (true metabolizable energy); TME n (true metabolizable energy corrected for nitrogen); VIF (variance inflation factor).

A moisture content prediction model for deep bed peanut drying using support vector regression

“…Furthermore, artificial neural network (ANN) had successfully described the drying kinetics of rice (Alam et al, 2018; Beigi, Torki‐Harchegani, & Tohidi, 2017), mosambi peel (Chaurasia, Younis, Qadri, Srivastava, & Osama, 2018) and quince (Taghinezhad, Kaveh, Jahanbakhshi, & Golpour, 2020). However, Faridi, Golian, Mottaghitalab, López, and France (2013) reported that the predictive ability of ANN models showed poor predictive performance on newly introduced data.…”

Bayesian analysis of energy balance data from growing cattle using parametric and non-parametric modelling

“…For instance, penalised and B-splines models have been used in zero-inflated Poisson models in animal abundance studies (Chiogna and Gaetan 2007), animal models in genetics (Cantet et al 2005), longitudinal nonparametric ANOVA models (Crainiceanu et al 2005) and random regression models in genetic analyses of cattle growth (Meyer 2005). Further, support vector regression was used by Faridi et al (2013) to predict the ME content of corn for ducks and neural network models were used by Faridi et al (2012) to evaluate egg production in response to dietary nutrient intake by hens. Non-parametric models often suffer from the lack of biological interpretation on parameters directly determining the shape of the response curve.…”