It is particularly important to test the quality of diesel oil. In order to improve the reliability and rapidity of the determination, a novel CC-tSNE-SVR model combined with near infrared spectroscopy (NIRS) was proposed for simultaneous determination of diesel multi parameter properties (density, viscosity and freezing point). The model was established by combining the method of correlation coefficient (CC), t-distributed stochastic neighbor embedding (tSNE) and support vector regression (SVR), which achieved the combination of wavelength selection algorithm and manifold learning technology. In order to test the validity of the method, several other models based on the same diesel data set of NIRS were compared. As a result, the CC-tSNE-SVR model presented in this paper not only had the best root mean square error of prediction (RMSE) and determinant coefficient (R2) in the analysis of diesel properties, but also had the fastest execution speed. Therefore, the combination of NIRS and the proposed CC-tSNE-SVR model open up a new way for the rapid and accurate monitoring of diesel quality.