The invention relates to an artificial neural network-based method for quickly forecasting transformation temperatures of Ni-Mn-Ga shape memory alloys. The method comprises the following steps of 1) collecting the Ni-Mn-Ga shape memory alloys and the transformation temperatures of the Ni-Mn-Ga shape memory alloys; 2) collecting component content of the Ni-Mn-Ga shape memory alloys; 3) by taking the component content as independent variables and the transformation temperatures as dependent variables, building a quick forecast model for the transformation temperatures of the Ni-Mn-Ga shape memory alloys by adopting an artificial neural network algorithm; and 4) according to the built quick forecast model for the transformation temperatures of the Ni-Mn-Ga shape memory alloys, and to-be-detected new Ni-Mn-Ga shape memory alloys and component content of the to-be-detected new Ni-Mn-Ga shape memory alloys, quickly forecasting the transformation temperatures of the Ni-Mn-Ga shape memory alloys. According to the method, an instrument does not need to be purchased; the operation is simple; and the cost is low. In the whole forecasting process, samples do not need to be chemically and physically treated, so that the samples are not destroyed; and no chemicals are used, so that environmental pollution is avoided.