This study aims to establish a method for estimating the composition of raw materials of eco-feed used in Okinawa prefecture, using near-infrared spectroscopy. We have developed and evaluated a calibration model and examined its practical feasibility. Herein, 100 samples were collected for calibration modeling and 20 for calibration model evaluation. Calibration models were constructed for evaluating general composition, gross energy (GE), and enzymatic digestibility of organic matter (EDOM) through multiple linear regression analysis (MLR) and partial least squares method (PLSR). The coefficients of determination (R2v) and standard deviations of the estimation error (SDP) for CP, EE, NFE and GE were 0.92 and 4.75, 0.90 and 3.52, 0.90 and 3.11, 0.86 and 0.27, respectively, through MLR. And R2v and SDP through PLSR were CP (0.96 and 3.15), EE (0.87 and 4.10), NFE (0.95 and 4.24), GE (0.86 and 0.27), respectively. For CP, EE, NFE and GE, the practical accuracy was evaluated very high or high. But for CA, CF and EDOM, there were no practical calibration models obtained through neither MLR nor PLSR.
View full abstract