A stepwise regression method with tropical cyclone intensity change as the dependent variable is developed on the basis of the climatology and persistence (CLIPER) forecast scheme. The predictors are involved in climatology persistence,synoptic predictors,infrared satellite data. Regression is performed on 24-,48-,72-hour forecasts for three different sea areas in Northwest Pacific. The climatology persistence predictors as important factors are chosen in regression equations, another predictors are some synoptic predictors with the dynamical forcing effect and the maximum potential intensity (MPI) which is calculated by sea surface temperature. The independence test by a jackknife procedure shows an improvement over CLIPER scheme. Infrared satellite predictors could improve the prediction accuracy on some forecast interval.