Simulated water samples of 3 kinds of preservatives and 4 kinds of sweeteners were formulated by using orthogonal design. Kernel independent component analysis (KICA) was used to process the UV spectra of the simulated water samples and the beverages added different amounts of the additive standards, then the independent components (ICs), i.e. the UV spectral profiles of the additives, and the ICs’ coefficient matrices were used to establish UV-KICA-SVR prediction model of the simulated preservatives and sweeteners solutions using support vector regression (SVR) analysis. The standards added beverages samples were obtained by adding different amounts level of additives in carbonated beverages, their UV spectra were processed by KICA, then IC information represented to the additives and other sample matrix were obtained, and the sample background can be deducted by removing the corresponding IC, other ICs’ coefficient matrices were used to estimate the amounts of the additives in the standard added beverage samples based on the UV-KICA-SVR model, while the intercept of linear regression equation of predicted amounts and the added amounts in the standard added samples is the additive content in the raw beverage sample. By utilization of chemometric “blind source separation” method for extracting IC information of the tested additives in the beverage and other sample matrix, and using SVR regression modeling to improve the traditional standard addition method, a new method was proposed for the screening of the preservatives and sweeteners in carbonated beverages.　The proposed UV-KICA-SVR method can be used to determine 3 kinds of preservatives and 4 kinds of sweetener in the carbonate beverages with the limit of detection (LOD) are located with the range 0.2～1.0 mg·L-1, which are comparable to that of the traditional high performance liquid chromatographic (HPLC) method.