Milled rice yield is one of the important indicators used to define the quality of milled rice. Near-infrared reflectance (NIR) spectroscopy was evaluated as a tool to estimate milled rice yield. Rough rice was dehusked and then milled to various milled rice yields. A total of 198 samples with milled rice yields of 85.72% to 96.91% were scanned over the NIR spectral wavelength ranging from 833 to 2500 nm with a Fourier transform near-infrared reflectance spectroscopy (FT-NIR) system. After finding the optimal spectral region (1638.8 to 2354.9 nm) according to the coefficients of determination, eleven mathematical pretreatments were performed on the selected spectra. The optimal calibration and prediction were obtained in the selected optimal spectral region using partial least square regression (PLS) on the spectra pretreated by a combination of first-derivation and multiplicative scattering correction. The best calibration model yielded a coefficient of determination (r2) of 0.994, a root mean square error of prediction (RMSEP) of 0.174%, and a bias of -0.021%, showing good predictability. Thus, the NIR spectroscopy technology has potential for predicting milled rice yield.