To screen disease which based on tongue inspection rapidly, the reflection spectrum on the tongue tips of 149 volunteers were collected by visible and near-infrared spectrometer and then the normalized reflectivity was calculated. Samples were divided into four classes according to the clinical diagnosis information: healthy, hyperviscosity, fatty liver, and coronary heart disease groups. Spectra were then subjected to three different analysis methods: principle component analysis (PCA) combined with artificial neural network (ANN), partial least squares (PLS), and interval PLS (iPLS). The classification accuracy of each model are 75%, 75%, and 85%, respectively. The results show that iPLS method sees more robust than the others. And the results also show that near-infrared region including more disease information than visible region. Experimental results show that the application of the spectra for disease diagnosis is promising.