Principal component analysis (PCA), partial least squares (PLS) regression and surface-enhanced Raman spectroscopy (SERS) are used in quantitative analysis of the concentration of fenitrothion solution. 600~1800 cm-1 SERS spectra of the fenitrothion solution are measured. The spectra around characteristic peaks are preprocessed by the absolute value of the first derivative, the multiplicative scatter correction (MSC) and the standard normal transformation respectively. Models are built by PCA and PLS regression. Grouping alternating method is used to validate the performance of models. It is found that the model built with the MSC preprocessed spectra performs better. The accuracy of analysis meets the detecting requirement of fenitrothion.