Stepwise regression (often called forward or backward regression) involves fitting a regression model and adding or removing predictors based on $t$ statistics, $R^2$ or information criteria to arrive in a *stepwise* manner at a final model. This tag can also be used for forward selection, backward elimination & best subsets variable selection strategies.
Note that performing inferential statistics via $$p$$-values after stepwise regression is invalid unless the $$p$$-values have been adjusted to account for the stepwise model building step.