The invention discloses a method for predicting flu antigen through a model and an application thereof. The method comprises the following steps: extracting 12 features affecting the flu antigen, wherein the 12 features include mutation numbers of five antigenic determinant amino acid, five physical and chemical properties of HA protein amino acid, factor affecting receptor, and number of glycosylated site, the five physical and chemical properties of amino acid include hydrophobicity, volume change, electriferous property, polarity and integrable surface area; and making statistics on the 12features of 3681 virus pairs similar to known antigen and 1720 virus pairs variant from antigen, thereby establishing a prediction model on antigen relation. On the basis of sequence, the method can present the antigen relation among viruses, and thus having the advantages of simpleness, convenience and high sensitivity. The process of antigen evolution can be visually displayed through the network. The invention is used for revealing flu spreading rules, screening vaccine candidate strains and the like.