In the present paper, UV-visible absorption spectrum and neural network theory were used for the analysis of cholesterol concentration. Experimental investigation shows that the absorption spectrum has the following characteristics in the wave band of 350-600 nm: (1) There is a stronger absorption peak at 416 nm for the test sample with different cholesterol concentration; (2) There is a shoulder peak between 450 and 500 nm, whose central wavelength is 460 nm; (3) There is a weaker peak at 578 nm; (4) Absorption spectrum's shape of different cholesterol concentration is different obviously. The absorption spectrum of serum is the synthesis result of cholesterol and other components (such as sugar), and the information is contained at each wavelength. There is no significant correlation between absorbance and cholesterol content at 416 nm, showing a random relation, so whether cholesterol content is abnormal is not determined by the absorbance peak at 416 nm. Based on the evident correlation between serum absorption spectrum and cholesterol concentration in the wave band of 455-475 nm, a neural network model was built to predict the cholesterol concentration. The correlation coefficient between predicted cholesterol content output A and objectives T reaches 0.968, which can be regarded as better prediction, and it provides a spectra test method of cholesterol concentration.