ABSTRACT: The nutritional content of feed ingredients is one of the factors considered in formulating livestock rations. Generally, the determination of nutritional content uses conventional methods, but these methods are destructive, expensive, and time-consuming, making them unsuitable for measuring nutritional content during ration formulation. This study aims to determine the accuracy of crude protein and crude fat content in rice bran using Artificial Neural Networks (ANN) based on NIRS absorbance data. This study used 60 rice bran samples from various regions representing West Sumatra. NIR spectral data were obtained using a Portable Fourier Transform Near Infrared (FT-NIR) device with a wavelength of 1000 nm-2500 nm. The results of the estimated nutritional content of rice bran were analyzed using an Artificial Neural Network (ANN) with 3, 5, 7, and 9 hidden nodes and 25,000, 30,000, 35,000, 40,000, and 50,000 iterations. The NIR absorbance data was pretreated by normalizing it using Unscrambler software and treating it using the PCA (Principal Component Analysis) method in IBM SPSS Statistics 21. The best estimation results can be seen in the lowest Standard Error of Prediction (SEP) and Coefficient of Variation (CV) values. The results showed that the use of JST with the developed model could estimate the crude protein and crude fat content of rice bran well and closely approximated the actual values. The crude protein estimation results have low SEP and CV values, namely SEP 1.26% and CV 14.91%, while the crude fat estimation results have SEP 1.21% and CV 15.12%.