Estimation of the TDN content of ear corn silage by neural network model

Abstract

In this study, we developed the models to predict the total digestible nutrients (TDN) content of ear corn silage (ECS) based on the contents of acid detergent lignin (ADL) and starch. The 51 samples of ECS were collected for the analysis of the chemical composition and the TDN content. The content of TDN was determined by the total feces collection method using wethers. The TDN content of the ECS ranged from 72.2 to 88.9% on the DM basis and averaged 79.5%. To explain the TDN content by the contents of both ADL and starch, we applied a multiple regression model or a neural network model (NNM). The multiple linear regression model showed low correlation between predicted and actual TDN contents (R²＝0.09). On the other hand, the NNM showed relatively high R² of 0.58 between predicted and actual TDN contents. These results indicated that the relationship among the contents of ADL, starch and TDN of the ECS should be nonlinear and the TDN content can be estimated more precisely by the NNM compared with the multiple regression model.