This paper proposes an entropy method for diversified fuzzy portfolio selection. Proportion entropy is introduced and credibilistic mean-variance and mean-semivariance diversification models for fuzzy portfolio selection are proposed. The crisp forms of the proposed models are also provided when the security returns are all triangular fuzzy variables. As an illustration, an application example of mean-variance diversification model is given using real data from Shanghai Stock Exchange. The computation results show that the proposed model results in a more diversified investment than the credibilistic mean-variance model.