Near-infrared (NIR) spectroscopic methods for measuring degradation products, including total polar materials (TPMs) and free fatty acids (FFAs), in soy-based frying oil used for frying various foods have been successfully developed. Calibration models were developed using forward stepwise multiple linear regression (FSMLR) and partial least-squares (PLS) regression techniques and then tested with an independent set of validation samples. The results show that the quality of oil used for frying different foods can be measured with a single model. First-derivative treatments improved results for TPM measurement. In addition, PLS models gave better prediction results than FSMLR models. For PLS models, the best correlations (r) between the NIR-predicted data and the chemical method data for TPMs and FFAs in oils were 0.995 and 0.981, respectively. For FSMLR models, the best r values for TPMs and FFAs in oils were 0.993 and 0.963, respectively.