A short-term weather prediction attempt is presented in this paper. Different from the traditional method, a prediction algorithm called addition-subtraction frequency (ASF) algorithm is applied to handling weather prediction. Practically, the snowy weather date data of Guangzhou are chosen as input to ASF algorithm. In order to make the prediction more accurate, three numerical experiments with different forms of input are conducted, and the forms of the input are year form, month/year form, and month/day/year form, respectively. On the basis of consistency analysis, years 2033, 2042 −2044, and 2052 are indicated as the most possible years of snowy winter in Guangzhou within a time span of forty years.