Abstract: In this paper a new modeling and forecasting scheme for ARMAX processes is proposed, which consists of three parts: pre-modeling, selftuning predictor (STP) and remodeling, with the STP as its principal part. Besides, the parameter increment correcting algorithm and the outlier detection method are presented in the scheme. The proposed scheme and algorithms can all be performed by employing linear least-square estimation algorithm, so that its computations are simple and fast. Simulations and practical applications show that this scheme has nice estimation and predication behaviours.