The feasibility of near infrared spectroscopy (NIRS) combined with stoichiometry for quantitative determination of umami substances and umami intensity in mixed solutions is discussed. The mixed solution of umami substance composed of monosodium glutamate and disodium inosinate is used as the research object to obtain samples of different umami intensity and collect their near-infrared spectral data. Based on the partial least squares regression method combined with the new competitive adaptive moving window interval combination (CMIC) algorithm, and a variety of commonly used variable optimization algorithms, the detection models of umami concentration and umami intensity in mixed solution are established. The experimental results show that the optimal detection models of umami concentration and umami intensity of mixed solution are simplified models based on CMIC algorithm, and the predictive determinations are 0.8886, 0.9182, and 0.8097, respectively. Therefore, the NIRS analysis technique combined with stoichiometry can be applied to quantitatively detect umami concentration and umami intensity in mixed solutions.