To further improve the ability of analyzing complex solution composition and extract more spectral information about absorption and scattering component, a multi-position and multi-optical path diffuse transmission spectroscopy is used to analyze complex solution composition quantitatively. 100 kinds of fixed solution containing different composition are formed in an experiment. The multi-position and multi-optical path transmission spectral distribution data of the near infrared light in the waveband from 650 nm to 1000 nm passing through the complex solution are simulated by using the Monte-Carlo (MC) method. A partial least square regression modeling method is used to model the spectral data at the central point in one optical path, the spectral data in multi-positions in one optical path and the spectral data in multi-positions and multi-optical paths with intralipid solution ingredients in the complex solution respectively. It is derived by comparison that the modeling accuracy and prediction of the composition in complex solution can be improved by increasing the spectral data in different radial positions and the spectral data in multi-optical paths.