The optimization of objective weighting method based on relative importance

The concept of “relative importance” is introduced to solve the problem of low weight differentiation obtained by objective weighting method. The experiment optimizes five objective weighting methods: grey relative analysis, mean square error method, coefficient of variance method, the entropy weight method and Gini coefficient method. The experiment resolves the index weight which is based on the relative importance between features. The experiment compared the difference of index weight between the original objective weighting method and the optimized objective weighting method. The conclusion is that the optimized grey relative analysis method and the optimized mean error method not only accord with the objective fact, but also improve the distinction between the weights of the indexes under the condition that the order of relative importance among the indexes remains the same.