School of Economics and Management,North China Electric Power University
Analyzing the characteristics of power material suppliers’ tender price is the basis for purchase enterprises to estimate tender price and formulate reasonable purchasing strategy. The traditional Spearman rank correlation coefficient feature selection method only considered the numerical order of features, and cannot mine the intrinsic relationship between feature variables and targets, which will lead to poor prediction effect. Copula function was introduced into the forecasting model of material tender price of power enterprises, and the dependent relationship between variables was analyzed through the joint probability distribution of characteristic variables and target variables. Firstly, the marginal distribution of characteristic variables and target variables were determined, and the copula parameters were estimated. Then, the appropriate copula function was selected and the correlation coefficient was calculated to screen the characteristics of the suppliers’ tender price. Finally, a variety of prediction methods were used to verify whether the prediction accuracy was improved after the introduction of Copula function. The results show that the copula function has higher prediction accuracy and better effect.