Based on daily heat index, the monthly heat index model was established using methods of the stepwise regression model, the gray model and the mean generating function model in every region of Jilin Province. The results showed that the average accuracy of the back substitution tests of three models were over 96%, the average ac-curacy of the extrapolate results of three models were over 94%, each model was able to forecast corn heat index preferably, and the mean generating function model was the best one.