BSTRACT
The preliminary study of using near infrared spectroscopy (NIRS) to predict crude protein in
shrimp feed was carried out. The calibration set of 41 shrimp feed samples was laboratorily prepared
against the validation set of 39 samples with similar distribution for the chemical content. The samples
were scanned for the NIR spectra in the wavelength region of 1100 nm to 2500 nm. Both multiple linear
regression (MLR) and partial least square regression (PLS) were performed to analyze data to create the
best calibration equations. The results indicated that absorbances at wavelengths of 1686 nm, 1778 nm
and 2486 nm were related with protein in the samples. At 1686 nm the absorbance was found to
coincide with that of protein analytically found the most in the prepared shrimp feed. Following the use
of wavelengths selected in MLR for constraining the wavelength region as input to PLS, the calibration
equations were shown to improve in prediction accuracy.