The invention provides a prediction method and system for a gamma globulin non-reactive Kawasaki disease. The method comprises the following steps of collecting 21 original parameters of an SVM model, wherein the original parameters for modeling include gender, age, fever time during doctor seeing, clinical classification, a CRP detection value, a WBC value, a PLT value, an Hb value, an ALT value, an AST value, an ALB value, gamma globulin usage time and clinical diagnosis symptom indexes, and the clinical diagnosis symptom indexes include conjunctival congestion, rash, cracked lips, strawberry-like tongue, neck lymphadenectasis, hand and foot scleredema, digit desquamation, perianal desquamation and vaccinated scar redness and swelling; performing discretization processing on the original parameters to obtain SVM eigenvalues corresponding to the original parameters; and building the SVM model by taking the SVM eigenvalues as basic data, and predicting gamma globulin non-reactive complications of the Kawasaki disease through the SVM model. According to the prediction method and system, patients can be subjected to early intervention treatment, thereby facilitating coronary artery injury recovery; and the prediction method and system has important significance and value for diagnosis and treatment of the Kawasaki disease.