Abstract: Traditional autoregressive moving-average model can’t be applied directly medium-term forecast on time series because noise occurs and accumulates with time. Aiming at this problem, the paper proposes adaptive autoregressive moving-average model, which will be divided into noise-resistant iterative model and noise observational model. According to the iterative model's characteristic, the study deduces detailed and gives perfectly a series of iterative solution formula to apply directly medium-term forecast on time series. Finally by predicting output values of Farming, animal husbandry and fishery (1985-2001) in Huang-Huai-Hai plain, it suggests that the model is reliable, and then output values of 2002-2020 are predicted to provide agricultural structural adjustment with the theory foundation.