To improve the quality of threshing and re-drying production, a method fused by random forest and orthogonal experiment scheme are presented to optimal the process parameters. Random forest performed regression modelling on the historical data of process parameters and moisture value, and ranked the process parameters. The model is verified by field test and the gap between the predict results and reality is less than 5%, which satisfied the on-site demands. To achieve optimal parameter combination with limited numbers of time of experiments, orthogonal experiment scheme according to the order of importance is designed. The optimal parameter combination was calculated according to the results and verified by the random forest model and field test. Results of our optimal combination in filed test showed it was better than the historical ones. This fusion method illustrates its potentials to promote the decision of process of threshing and re-drying.