This paper proposes an indices' screening and weighting method based on the information sensitivity to solve the reasonably indices' screening and weighting problem for the existing research. This paper has two main innovation points. One is multiplying the proportion of variance by the partial derivative each retained principal component sub index, and use their sum to indicate effect degree of this index on the information of the original indicators set. The way of screening indicators is retaining the index for the larger information sensitivity. The indices' accumulated information contents are calculated by means of the information sensitivity screening standard. And then, the indices are screened out based on the size of these. Two is calculating the percent of an index to all indices' information sensitivity, take it as the weights of this index to reflect relative information content. Finally, the proposed method and the existing principal components method are contrastive analyzed by a numerical illustration, on this account the feasibility and validity of the proposed method are illustrated.