This paper presented an analysis method of TFAs in hydrogenated soybean oil based on THz spectroscopy. The THz time domain spectrum of hydrogenated soybean oil was transformed into frequency domain spectrum by Fourier transform， and the refractive index and absorption spectra were calculated by using the optical parameters. In order to study the absorption characteristics of TFAs in hydrogenated soybean oil in terahertz band， the RB3LYP in DFT and 6-311G (d， p) were adopted for optimizing the geometric configuration and calculating vibration frequency of the several major trans oleic acids and trans linoleic acids in hydrogenated soybean oil. These studies laid a theoretical basis for the composition and qualitative analysis of TFAs in hydrogenated soybean oil. The prediction model of TFAs content in hydrogenated soybean oil was established by LS-SVR， and compared with PLS and SVR models. The prediction results demonstrated that the LS-SVR model was the best one. The RMSEP was 0.3246， the determination coefficient R2 was 0.9792， and the RSD was 3.81%， which could meet the actual detection requirements. Then， PSO algorithm was used to optimize the parameters of LS-SVR model， and compared with GS and GA optimization methods， the optimal parameter combination was selected. The RMSEP， R2 and RSD of the optimized LS-SVR model were 0.0763， 0.9989 and 0.90%， respectively， which significantly improved the prediction accuracy. This study laid a foundation for the rapid detection of TFAs in hydrogenated soybean oil by terahertz spectroscopy.