Abstract: The point prediction problem of the unknown random variable under asymmetric loss function is discussed within the framework of decisio n theoretic. By introducing an improving equivariant estimator method in the est imation theory into statistical prediction problem, the usual equivariant predic tor can be improved. By means of the order restriction of the unknown distributi on parameters, the best equivariant predictor is improved and a class of imp roved predictors are constructed under some conditions. Some examples are given.

Key words: point prediction, LINEX loss, order restriction, equiv ariant predictor, improved predictor